www.sciencemag.org/content/363/6427/644/suppl/DC1

Supplementary Materials for

Tumor metastasis to lymph nodes requires YAP-dependent metabolic adaptation

Choong-kun Lee, Seung-hwan Jeong, Cholsoon Jang, Hosung Bae, Yoo Hyung Kim, Intae Park, Sang Kyum Kim, Gou Young Koh*

*Corresponding author. Email: gykoh@kaist.ac.kr

Published 8 February 2019, Science 363, 644 (2019)

DOI: 10.1126/science.aav0173

This PDF file includes:

Materials and Methods Figs. S1 to S13 Table S1 References

2

Materials and Methods

Mice, generations of tumor models, and etomoxir treatment for FAO inhibition

All animal studies were performed according to the protocols (KA2017-10) approved by the

Animal Care Committee of the Korea Advanced Institute of Science and Technology (KAIST).

Specific pathogen-free C57BL/6J, BALB/c and MMTV-PyMT transgenic mice (FVB/N) were

purchased from Jackson Laboratory (Jackson Labs, ME). All mice were fed with ad libitum

access to standard diet (PMI Lab diet) and water, and were anesthetized by intra-peritoneal

injection of a combination of anesthetics (80 mg/kg of ketamine and 12 mg/kg of xylazine)

before all surgical and sampling procedures. B16F10 melanoma cells were purchased from

American Type Culture Collection and GFP+ B16F10 (B16F10-GFP) and GFP+ 4T1 (4T1-GFP;

4T1) breast cancer cells were purchased from Imanis Life Sciences. For B16F10 cell footpad

implantation model, 5 x 105 B16F10 or B16F10-GFP cells were implanted subcutaneously into

the footpad region of hind limb of 8~9-week-old male C57BL/6J mice. Primary tumor volume

was monitored and calculated according to the formula 0.5 x A x B2, where A is the largest

diameter of a tumor and B is its perpendicular diameter. Popliteal LNs (pLNs) as mapped by

Harrell and colleagues (29) were dissected for further experiments and analyses. For FAO

inhibition, etomoxir (40 mg/kg of body weight, Sigma-Aldrich) in PBS (50 μl) was injected daily

subcutaneously to the anterolateral side of mouse leg, right above the hind limb foot where

B16F10 cells were implanted. Delivery to pLN by this administration route was confirmed by

Evans blue dye in a separate experiment. The treatment was started from the 2nd week after

tumor implantation. Ipsilateral pLNs and primary tumor tissues were sampled at day 25. For

orthotopic 4T1 breast cancer model, 1 x 106 4T1 cells were implanted into the third mammary fat

pad on the right side of 8-week-old female BALB/c mice. Etomoxir (40 mg/kg of body weight)

was injected daily subcutaneously into the footpad of forelimb for delivery to axillary LNs. The

treatment began from the second week after tumor implantation. Ipsilateral axillary LNs were

sampled at day 25. For MMTV-PyMT spontaneous breast cancer model, etomoxir (40 mg/kg of

body weight) was injected every other day into the footpad for delivery to axillary LNs. The

treatment began when the mice were 13 weeks old, and the primary tumor and ipsilateral axillary

LNs were sampled 3 weeks later. To generate a blood-borne lung metastasis model, 1 x 106

B16F10 or 4T1 cells in 100 μl of PBS were injected into tail vein of 8~9-week-old male

C57BL/6J or female BALB/c mice, respectively. Etomoxir (40 mg/kg of body weight) in PBS

(50 μl) was injected daily into tail vein and retro-orbital venous sinus to examine the effect of

FAO inhibition in the blood-borne lung metastasis model. The lungs with metastatic tumor

nodules were sampled one week later. For direct implantation of tumor cells into LNs, 5 x 104

B16F10 or 4T1 cells in 10 μl of PBS were injected into inguinal LNs of male C57BL/6J or

female BALB/c mice, respectively, using a 30 gauge syringe. The inguinal LNs were sampled at

day 3 after implantation. For removal of primary tumor, the tumor masses were surgically

dissected under sterile condition at the second week after implantation of the B16F10 cells into

the footpad. Wound was sutured with 6-0 black silk after hemostasis by manual compression for

5 min. To detect migrating tumor cells in lymphatic vessels, 2 x 105 B16F10-GFP cells in 50%

Matrigel (BD Biosciences) were implanted to the tip of the ear of C57BL/6J mice. The ears were

sampled for whole-mount staining 5 days later. To knock down YAP in the implanted B16F10

melanoma, doxycycline diet (0.625 mg/kg) (30) was given from the second week after the

implantation. To see the effects of bile acid on LN metastatic tumors, taurodeoxycholic acid (3.2

mg/kg, Sigma-Aldrich) in PBS (50 μl) was injected subcutaneously every day to the anterolateral

3

side of mouse leg, right above the hind limb foot where B16F10 melanoma was implanted

according to the indicated schedule.

RNA-sequencing (RNA-seq) and bioinformatics data analysis

B16F10-GFP+ tumor cells from primary site (footpad) and pLNs (isolated at 3 weeks after

implantation for micro-metastatic tumor with metastatic foci diameter less than 2 mm and at 4

weeks after implantation for macro-metastatic tumor) were isolated. Primary tumor and pooled

pLNs were digested in 5 ml of enzyme buffer containing 0.2 mg/ml collagenase type-II

(Worthington), 0.1 mg/ml DNase I (Roche) and 0.8 mg/ml dispase (Gibco) at 37°C for 30 min.

The tissues were gently agitated every 5 min during digestion to disrupt any cell clumps. After

complete digestion, the cell suspension was filtered through 40-μm nylon cell strainer and

washed. After RBC lysis with suspension in the ACK lysis buffer (Gibco) for 2 min at room

temperature (RT), the cells were incubated for 15 min with anti-CD45 microbeads (Miltenyi

Biotec) to deplete CD45+ hematopoietic cells using AutoMACS (Miltenyi Biotec) according to

the manufacturer’s instructions. The GFP+ tumor cells were sorted by FACS Aria II (BD

Bioscience) from the rest of enriched tumor cell fractions. The cDNA library was constructed

using SENSE 3’ mRNA-Seq Library Prep Kit (Lexogen, Inc.) according to the manufacturer’s

instruction. The library was amplified to add the complete adapter sequences required for cluster

generation. High-throughput sequencing was performed as single-end 75 sequencing using

NextSeq 500 (Illumina, Inc.). Mapping of RNA-Seq reads were performed using Bowtie2

version 2.1.0. The alignment file was used to assemble transcripts, estimate their abundances,

and detect differential expression of genes or isoforms using cufflinks. The Read Count data

were processed based on Quantile normalization method using the Genowiz ™ version 4.0.5.6

(Ocimum Biosolutions, India). For bioinformatics analysis, each gene expression data was

annotated with official gene symbols, then normalized (log2). The principal component analysis

(PCA) data were analyzed using Genowiz™ version 4.0.5.6 (Ocimum Biosolutions, India).

Unsupervised hierarchically clustered heatmaps were generated by using GeneCluster 3.0

(University of Tokyo, Human Genome Center). Gene set enrichment analysis (31) (GSEA) was

performed with hallmark (h) or oncogenic (c6) gene set collections of the Molecular Signature

Database v6.1 (http://www.broadinstitute.org/gsea/msigdb). Gene classification analysis and

upstream regulator analysis was performed with Ingenuity Pathway Analysis (IPA, Qiagen).

Collection of lymph fluids from mice

Lymph fluid was collected from the mice according to a previously described method (32).

Briefly, the mice were anesthetized with isoflurane in oxygen via a face mask and positioned in

right-decubitus position under the microsurgery system. The thoracic duct behind the aorta was

exposed by opening of the left side of abdomen and dissecting the dorsal parietal peritoneum.

Cannulation with 20-gauge polyethylene tube was performed into the thoracic duct, and lymph

fluid was collected into the EDTA-coated tube. The collected lymph fluid was immediately

frozen in liquid nitrogen and stored at -80°C until use.

Metabolomics analysis

Tissue samples from footpad, naïve pLN, B16F10 primary tumor, and B16F10 LN metastatic

tumor, and blood or lymph fluid samples from tumor-naïve or LN metastatic tumor-bearing adult

C57BL/6J mice were prepared for metabolome analyses. Sampling of tumor tissue, blood and

lymph was conducted 4 weeks after implantation of B16F10 cells into the footpad. The tumor

4

tissue in the LNs was carefully dissected under microscope. The samples were immediately

frozen in liquid nitrogen and stored at -80°C until sent to Human Metabolome Technologies

(HMT, Yamagata, Japan). The transferred samples were prepared and analyzed by the two

modes (for detection of both cationic and anionic metabolites) of capillary electrophoresis time-

of-flight mass spectrometry, liquid chromatography time-of-flight mass spectrometry. The

samples of blood and lymph fluid were analyzed by liquid chromatography-tandem mass

spectrometry. Identification of metabolites from the peaks was based on the annotated tables of

m/z values and normalized migration times. Relative peak areas under the curves were

quantified. Although the peaks of two muricholate isomers (muricholic acids, MCA) were

detected in the samples, the metabolomics analysis was not capable of identifying the different

MCA isoforms due to a technical limitation.

Metabolic assays

To measure fatty acid uptake in vivo, a bolus dose of 2.0 mCi of 14C-oleic acid (Perkin Elmer)

dissolved in 200 µl of olive oil was given by oral gavage to the mice bearing primary and LN

metastatic B16F10 melanomas (at 4 weeks after implantation of B16F10 cells into footpad).

After 2 hr, primary tumor and LN metastatic tumors were dissected and harvested. The tumors

were dissolved in 1 ml of tissue solubilizer (Solvable, Perkin Elmer) for overnight at 50°C,

decolorized with 0.3 ml of 30 % hydrogen peroxide for 1 hr at room temperature. Scintillation

solution (Ultima Gold, Perkin Elmer) was added to each vial. Total radioactivity was measured

by liquid scintillation using a Tri-Carb 2910TR Liquid Scintillatior (Perkin Elmer). Amount of 14C-oleic acid was normalized to mg of tumor tissue. We adapted previous methods (33, 34) to

measure ex vivo oxidation of 14C-palmitic acid (labeled in carbon 1, Perkin Elmer) in the

homogenates of tumor tissues. The homogenates were incubated in reaction mixture (33)

containing 0.4 µCi of 14C-palmitic acid for 30 min. Then 20 μl of 1 M NaOH was added to round

discs of Whatman filter paper inside the cap of reaction tube, 200 μl of 1 M perchloric acid was

subsequently added into the reaction mixture of the tube, and incubated for additional 1 hr at

room temperature. 14C radioactivity derived from the 14C-acid-soluble metabolites and the 14CO2

trapped by the paper discs were measured using Tri-Carb 2910TR Liquid Scintillation Analyzer

(Perkin Elmer). Measured 14C radioactivity was normalized to the tissue protein amount. To

measure the relative FAO activity in vitro, tumor cells were seeded in 24-well plates at 100,000

cells/well in triplicate, incubated with 14C-palmitic acid at 0.2 µCi per well for 3 hr at 37°C, and

trapped 14CO2 were measured. Measured 14C radioactivity was normalized to the cell protein

amount. OCR was measured using a Seahorse XFe96 analyzer and Mito Fuel Flex Test Kit

(Agilent) following the manufacturer’s instruction. Three major fuel pathway inhibitors,

mitochondrial pyruvate carrier inhibitor UK5099, glutaminase inhibitor BPTES and carnitine

palmitoyltransferase 1 inhibitor etomoxir, were sequentially treated to the on 96 well plates.

Fuel dependency was calculated as;

(baseline OCR - target inhibitor OCR)/(baseline OCR - all inhibitors OCR)

Fuel capacity was calculated as;

1 - [(baseline OCR - other 2 inhibitors OCR)/(baseline OCR - all inhibitors OCR)].

To detect radiolabeled cholesterol conversion into bile acids, the adapted B16F10 cells were

incubated in medium containing 14C-cholesterol (Perkin Elmer) substrate for 16 hr and

additionally incubated in fresh medium for 16 hr. Then the cells were harvested with -80°C

5

chilled methanol, and the supernatants were obtained by centrifugation at 14,000 rpm for 10 min

and then analyzed by thin-layer chromatography (TLC). The TLC plate (silica-gel) was

developed in isoamyl acetate:propionic acid:1-propanol:water (4:3:2:1) solvent system. The

radioactive materials on the TLC plate were visualized by autoradiography (Typhoon FLA 7000,

GE Healthcare). Mobility of the radiolabeled bands was compared with the mobility of bile acid

standards, visualized by iodine staining.

Generation of the metastasis-prone adapted B16F10 cells

The metastasis-prone adapted B16F10 cells were generated according to previous method

(35). Briefly, the metastatic pLN from the implantation of B16F10-GFP cells into footpad mouse

model was harvested and digested as aforementioned, and the tumor cells were selected in vitro

by addition of puromycin (1 µg/ml). Then, the expanded tumor cells were re-implanted to the

mice footpad. After three rounds of in vivo selection, highly LN metastatic B16F10 sub-

population was established, named as “adapted B16F10 cells”. The “parental B16F10 cells” was

established in the cultured cells that were isolated from primary implanted B16F10 melanoma of

the mice.

Histological analyses

For immunofluorescence (IF) staining, primary tumors, LNs and lungs were fixed in 1 % PFA

overnight, dehydrated in 20 % sucrose solution, and embedded in tissue freezing medium

(Leica). Frozen blocks were cut into 10 μm sections. Mid-sectioned tissues and organs were used

for IF staining. For whole-mount IF staining of the ear skin, shaved ear skins were fixed in 4%

PFA for 1 hr, and washed with PBS. Samples were blocked with 5 % goat (or donkey) serum in

PBST (0.3% Triton X-100 in PBS) and then incubated at 4°C overnight with the following

primary antibodies (all 1:200): anti-YAP (rabbit, clone D8H1X, Cell Signaling Technology),

anti-CD31 (hamster, clone 2H8, Millipore), anti-PDGFRβ (rat, eBioscience), anti-Ki-67 (rabbit,

Abcam), anti-LYVE-1 (rabbit, Angiobio), anti-GFP (chicken, Abcam), anti-S100 (rabbit,

Biocare Medical), anti-CK8 (rat, DSHB), and anti-GLUT1 (rabbit, Millipore). After several

washes, the samples were incubated for 2 hr at RT with the following secondary antibodies (all

1:1000): Alexa Fluor 647-conjugated anti-hamster IgG (Jackson ImmunoResearch), Alexa Fluor

488-, or 594-conjugated anti-rabbit IgG (Jackson ImmunoResearch), Alexa Fluor 488-, or 594-

conjugated anti-rat IgG (Jackson ImmunoResearch), and Alexa Fluor 488-conjugated anti-

chicken IgG (Jackson ImmunoResearch). Nuclei were stained with 4,6-diamidino-2-

phenylindole (DAPI, Invitrogen). Then the samples were mounted with fluorescent mounting

medium (DAKO) and immunofluorescent images were acquired using an LSM800 or LSM880

confocal microscope (Carl Zeiss).

Morphometric analyses

Morphometric analyses of LNs were performed with mid-sectioned images by analysis using

ImageJ software (http://rsb.info.nih.gov/ij) or ZEN 2012 software (Carl Zeiss). Nuclear

localization of YAP in tumor cells was determined by counting the tumor cells with YAP+

signals co-localized with DAPI signal in random three or four fields of primary or LN metastatic

tumors and averaged. The invasive front was defined as the region within 100 μm from the edge

of the tumor within LN stroma. To estimate the metastatic area in LN of B16F10 melanoma,

black-pigmented melanoma area from gross images was analyzed using ImageJ and expressed as

a percentage of total LN area. The measurement of metastasized breast cancer cells in LNs were

6

done in the mid-section area. The area of CK8+ fluorescence (MMTV-PyMT) or GFP+

fluorescence (4T1-GFP) was presented as a percentage of total mid-section area of LN. The

measurement of metastasized cancer cells in lungs were performed with hematoxylin and eosin

(H&E) stained images of lung sections acquired with Axio Zoom V16 (Zeiss). Three random

field of views per section and 3 sections per sample were analyzed. Metastatic area fraction was

analyzed using ImageJ image processing software and expressed as a percentage of lung

cellularity area.

Cell culture and treatments, and cell viability assay

The cancer cells were maintained in Dulbecco’s modified Eagle medium (DMEM)

supplemented with 10 % fetal bovine serum, penicillin (100 U/ml), and streptomycin (100

mg/ml). For serum starvation, the cells were incubated in DMEM without these supplements.

B16F10 (adapted) or 4T1 cells were transfected with siRNAs using Lipofectamine RNAiMAX

(Invitrogen) according to the manufacturer’s protocol. The pre-designed siRNAs targeting

following genes were used: Myc (No. 1389366), Egfr (No. 1353550), Kras (No. 1377654), Akt1

(No. 1320627), Yap1 (No. 148744), Cyp7a1 (#1 No. 1345396, #2 No. 1345397), Fxr (No.

20186-1), Pxr (No. 18171-1), Vdr (No. 22337-1), Iqgap1 (No. 29875-1), S1pr2 (No. 14739-1)

(Bioneer, Daejeon, Korea) and negative control siRNA (5’-GTCAAACGCCTTGTATTTA-3’).

Cells were harvested 24-48 hr after transfection. Efficiency of knockdown was evaluated by

qRT-PCR or immunoblotting. Doxycyline-inducible YAP knockdown or control B16F10 cells

were generated using a SMARTvector Lentiviral shRNAs (GE Dharmacon) according to the

manufacturer’s protocol. Water-soluble cholesterol, taurocholic acid (TCA), taurodeoxycholic

acid (TDCA), methyl-β-cyclodextrin and other chemicals were purchased from Sigma-Aldrich.

To investigate the effect of bile acid or cholesterol on YAP activation, adapted B16F10 cells

were serum-starved overnight and then bile acid (TDCA or TCA) or cholesterol was treated for

the indicated time. The doses for in vitro bile acids or cholesterol treatments were chosen by

referring to previous studies (36-39). Cell viability assay was performed according to the

manufacturer’s instructions for Cell Counting Kit-8 (CCK-8, Dojindo Molecular Technologies).

The adapted B16F10 cells were collected at the exponential phase and plated onto a 96-well

plate at a density of 5,000 cells/well. After incubation for 18 hr in DMEM, each well was

replaced with 100 μl of low glucose media supplemented with fatty acid (palmitic acid)

conjugated with BSA (FA-BSA, at final concentration of 0.2 mM), low glucose media

supplemented with BSA or high glucose media supplemented with BSA and then incubated for 8

hr. Each group was then treated with different concentrations of etomoxir from 0 to 200 μM and

incubated for 18 hr. Finally, the wells were incubated for 4 hr with 10 μl of the CCK-8 solution

followed by measurement of the absorbance at 450 nm using a microplate reader (Model 680,

Bio-Rad).

Retroviral infection

FLAG-YAP-5SA were cloned into retroviral pMSCV-puro vectors as previously described

(40). For retroviral particle assembly, retroviral constructs (pMSCV-puro, pMSCV-puro FLAG-

YAP-5SA) and two packaging plasmids (pCMV-VSV-G and pCMV-Gag-Pol) were co-

transfected to HEK293T cells using Lipofectamine LTX and PLUS reagent (Invitrogen), and the

retroviral supernatant was collected during 24-72 hr after transfection. Viral supernatant were

cleared by centrifugation at 3,000 rpm for 10 min, supplemented with 8 μg/ml polybrene (Sigma-

7

Aldrich), and infected to target cells. Cells were selected with puromycin (3 μg/ml) in culture

medium.

Quantitative Real-Time PCR

Total RNA was extracted from the cultured cells and tumors, and the liver and skeletal muscle

of healthy normal C57BL/6J mice using Trizol RNA extraction kit (Invitrogen) according to the

manufacturer’s instructions. Total RNA was reverse transcribed into cDNA using GoScriptTM

Reverse Transcription Kit (Promega). Then, quantitative real-time PCR was performed using

FastStart SYBR Green Master mix (Roche) and S1000 Thermal Cycler (Bio-Rad). GAPDH was

used as a reference gene and the results were presented as relative expressions to control. PCR

primer sequences specific for each gene are listed below (5’-3’): Gapdh (forward), TGT TCC

TAC CCC CAA TGT GT; Gapdh (reverse), TGT GAG GGA GAT GCT CAG TG; Yap1

(forward), GTC CTC CTT TGA GAT CCC TGA; Yap1 (reverse), TGT TGT TGT CTG ATC

GTT GTG AT; Myc (forward), CGG ACA CAC AAC GTC TTG GAA; Myc (reverse), AGG

ATG TAG GCG GTG GCT TTT; Egfr (forward), GAA AAG AAA GTC TGC CAA GGC ACA

AG; Egfr (reverse), GTA ATT TCC AAG TTC CCA AGG ACC AC; Kras (forward), CAA

GAG CGC CTT GAC GAT ACA; Kras (reverse), CCA AGA GAC AGG TTT CTC CAT C;

Akt1 (forward), GGC TGG CTG CAC AAA CG; Akt1 (reverse), GAC TCT CGC TGA TCC

ACA TCC T; Ctgf (forward), GTC CCA GAA CGC ACA CTG; Ctgf (reverse), CCC CGG TTA

CAC TCC AAA; Cyr61 (forward), GGA TCT GTG AAG TGC GTC CT; Cyr61 (reverse), CTG

CAT TTC TTG CCC TTT TT; Ankrd1 (forward), GCT GGT AAC AGG CAA AAA GAA C;

Ankrd1 (reverse), CCT CTC GCA GTT TCT CGC T; Iqgap1 (forward), ACT CGG CGC TGA

ACT CTA AG; Iqgap1 (reverse), TTT GGA GAC CTT GCC CTT GC; S1pr2 (forward), ATC

TAT ACG TGG CGT AGC CG; S1pr2 (reverse), GAC CAC TGT GTT ACC CTC CAG; Fxr

(forward), TGT TTC TTC GTT CGG CGG AG; Fxr (reverse), GAG AGA GGA TGA CGA

TCG CTG; Vdr (forward), CAT TGT CTC CCC AGA CCG AC; Vdr (reverse), GGA TAG

GCG GTC CTG AAT GG; Pxr (forward), CTC CCT CTT CTC CCC AGA TCG; and Pxr

(reverse), GTG AGC AGG ATA TGG CCG AC.

Immunoblotting.

For immunoblot analysis, harvested cells and homogenized tumor tissues were lysed on ice in

RIPA lysis buffer supplemented with protease and phosphatase inhibitors (Cell Signaling

Technology, #5872). The lysates were centrifuged for 10 min at 4°C, 13,000 rpm. Protein

concentration of the supernatants was quantitated using the detergent-insensitive Pierce BCA

protein assay kit (Thermo Scientific, 23227). Lamni buffer was added to total protein lysates and

samples were denatured at 95°C for 5 min. Aliquots of each protein lysate (10-20 μg) was

subjected to SDS–polyacrylamide gel electrophoresis. After electrophoresis, proteins were

transferred to nitrocellulose membranes and blocked for 30 min with 3 % BSA in TBST (0.1 %

Tween 20 in TBS). Primary antibodies were incubated overnight at 4ºC. After washes,

membranes were incubated with anti-rabbit peroxidase-coupled antibody (1:10,000, Cell

Signaling Technology, #7074) for 1 hr at RT. Target proteins were detected using ECL western

blot detection solution (Millipore, WBKLS0500). Blot intensity was measured with ImageJ.

Primary antibodies used for immunoblotting were as follows: rabbit anti-phospho-YAP (at

Ser127) polyclonal (1:1000, Cell Signaling Technology, #4911); rabbit anti-YAP monoclonal

(1:1000, Cell Signaling Technology, #14074); rabbit anti-β-actin monoclonal (1:2000, Sigma-

Aldrich, A5441); rabbit anti-CYP7A1 polyclonal (1:1000, BIOSS, bs-2399R).

8

Analyses of human melanoma LNs

To examine YAP distribution in human sentinel LN metastatic tumor and its relation to

prognosis, the sentinel LNs from melanoma patients who underwent LN dissection during the

surgical removal of primary tumor were retrieved according to the protocol approved by the

Institutional Review Board at Severance Hospital, Korea (IRB No.: 4-2018-0469). Among 73

formalin-fixed, paraffin-embedded sentinel LNs, metastatic tumors were found in 21 cases. IHC

was performed with anti-YAP antibody (sc-10119, lot no. J1318, Santa Cruz Biotechnology)

using an automated staining instrument (Ventana BenchMark XT, Ventana Medical System,

Inc.). Ultraview Universal Alkaline Phosphatase Red Detection Kit (Roche Diagnostics) was

used for detection of immune-positive signal. All IHC-stained slides were interpreted by a

blinded expert pathologist. YAP localization was classified as nuclear or cytoplasmic according

to the subcellular YAP expression. Patient demographics and clinical characteristics were

obtained from electronic medical records. Distant metastasis-free survival was defined as the

time from surgery to the date of diagnosis of distant organ metastasis.

Statistical analyses

Data are presented as mean ± standard error of the mean (s.e.m.). Reproducibility was ensured

by performing more than three independent experiments. Statistical differences between the

means were compared by the two-tailed, unpaired t-test for two groups, or determined using one-

way ANOVA followed by Tukey’s multiple comparison test for multiple groups, unless

otherwise noted. Statistical analysis was performed with PASW Statistics 18 (SPSS) or Prism 7

(GraphPad). The statistical software package R version 3.2.1 was used to run multivariate

analysis (Cox proportional-hazards regression model). Similarity between groups in PCA space

was compared by pairwise inter-cluster Mahalanobis distance. Mahalanobis distance between

groups (PT - Micro, PT - Macro) was calculated using covariance of primary tumor group.

Distances were calculated using MATLAB with custom-written software (MathWorks).

Statistical significance was set to P value less than 0.05. Statistical information are presented

below.

Fig. 2A. P = 0.0036 by two-tailed unpaired t-test. Fig. 2B. P = 0.0076 (glucose vs fatty acid) and

P < 0.0001 (glutamine vs fatty acid) by one-way ANOVA followed by Tukey’s multiple

comparison test. Fig. 2C. P = 0.0015 (PT vs LMT), P = 0.0004 (PT vs BAT), P = 0.0024 (Lung-

MT vs LMT), and P = 0.0006 (Lung-MT vs BAT) by one-way ANOVA followed by Tukey’s

multiple comparison test. Fig. 2F. P < 0.0001 by Fisher’s exact test for LN metastasis

percentage. P = 0.0036 by two-tailed unpaired t-test for metastatic tumor area in LN. Fig. 3A. P

= 0.0002 (siCtrl vs siYAP) by one-way ANOVA followed by Tukey’s multiple comparison test.

Fig. 3B. P = 0.0053 (glucose vs fatty acid) and P = 0.0192 (glutamine vs fatty acid) by one-way

ANOVA followed by Tukey’s multiple comparison test. Fig. 3I. P < 0.0001 (B16F10) and P <

0.0001 (MMTV-PyMT) by two-tailed unpaired t-test. Fig. 3M. P = 0.0411 by two-tailed unpaired

t-test for metastatic tumor area in LN. Fig. 4A. P = 0.0043 (LN metastatic tumor for CA), P =

0.0005 (LN metastatic tumor for DCA), P < 0.0001 (LN metastatic tumor for TCA), P < 0.0001

(LN metastatic tumor for TDCA), P = 0.0130 (Primary tumor vs LN metastatic tumor, MCA#1),

and P = 0.0059 (Primary tumor vs LN metastatic tumor, MCA#2) by one-way ANOVA followed

by Tukey’s multiple comparison test. Fig. 4G. P = 0.0037 by Log-rank test.

9

Supplementary Fig. S1D. P < 0.0001 by two-tailed unpaired t-test. Supplementary Fig. S2C.

P = 0.0343 (primary tumor volume) and P = 0.0070 (metastatic tumor area in LN) by two-tailed

unpaired t-test. Supplementary Fig. S2E. P = 0.0035 by two-tailed unpaired t-test.

Supplementary Fig. S3B. P = 0.0030 (PT vs LMT), P < 0.0001 (PT vs BAT), P = 0.0082 (LMT

vs BAT), P = 0.0074 (Lung-MT vs LMT), and P < 0.0001 (Lung-MT vs BAT) by one-way

ANOVA followed by Tukey’s multiple comparison test. Supplementary Fig. S3C. P = 0.0037

(PT vs LMT) and P = 0.0333 (Naïve LN vs LMT) by one-way ANOVA followed by Tukey’s

multiple comparison test. Supplementary Fig. S4G. P = 0.0152 by Fisher’s exact test for LN

metastasis rate and P = 0.0027 by two-tailed unpaired t-test for metastatic tumor area in LN.

Supplementary Fig. S4J. P = 0.0014 by Fisher’s exact test for LN metastasis rate and P =

0.0018 by two-tailed unpaired t-test for metastatic tumor area in LN. Supplementary Fig. S4M.

P = 0.0294 by Fisher’s exact test for LN metastasis rate and P = 0.0213 by two-tailed unpaired t-

test for metastatic tumor area in LN. Supplementary Fig. S5C. P = 0.0031 by Fisher’s exact test

for LN metastatic foci formation and P = 0.0007 by two-tailed unpaired t-test for metastatic

tumor area in LN. Supplementary Fig. S5E. P = 0.0012 by two-tailed unpaired t-test.

Supplementary Fig. S6B. P = 0.0006 (siMYC), P = 0.0196 (siKRAS), P = 0.0190 (siAKT), and

P = 0.0154 (siYAP) by two-tailed unpaired t-test. Supplementary Fig. S6C. P = 0.0103 (Ctgf),

P < 0.0001 (Cyr61), and P < 0.0001 (Ankrd1) by two-tailed unpaired t-test. Supplementary Fig.

S6D. P = 0.0009 by two-tailed unpaired t-test. Supplementary Fig. S6E. P = 0.0487 by two-

tailed unpaired t-test. Supplementary Fig. S7D. P = 0.0044 by two-tailed unpaired t-test.

Supplementary Fig. S7E. P < 0.0001 by two-tailed unpaired t-test. Supplementary Fig. S8C. P

= 0.0445 by two-tailed unpaired t-test for metastatic tumor area in LN. Supplementary Fig.

S8F. P = 0.0023 by Fisher’s exact test for LN metastatic foci formation and P = 0.0014 by two-

tailed unpaired t-test for metastatic tumor area in LN. Supplementary Fig. S8M. P = 0.0008 by

two-tailed unpaired t-test. Supplementary Fig. S9F. P = 0.0010 (30 min vs 0 min for TDCA) by

one-way ANOVA followed by Tukey’s multiple comparison test. Supplementary Fig. S9L. P =

0.0460 by two-tailed unpaired t-test for tumor metastatic area. Supplementary Fig. S10B. P =

0.0004 (60 min vs 0 min for cholesterol) and P = 0.0007 (90 min vs 0 min for cholesterol) by

one-way ANOVA followed by Tukey’s multiple comparison test. Supplementary Fig. S10C. P

= 0.0035 (siCYP7A1 #1 vs siCtrl) and P = 0.0015 (siCYP7A1 #2 vs siCtrl) by one-way ANOVA

followed by Tukey’s multiple comparison test. Supplementary Fig. S10D. P = 0.0135 by two-

tailed unpaired t-test. Supplementary Fig. S10F. P = 0.0226 (siCtrl vs siCYP7A1, Chol), P =

0.0117 (siCtrl vs siCYP7A1, 7αOHChol), and P = 0.0499 (siCtrl vs siCYP7A1, CDCA) by two-

tailed unpaired t-test. Supplementary Fig. S10I. P = 0.0015 by two-tailed unpaired t-test.

Supplementary Fig. S10K. P < 0.0001 by two-tailed unpaired t-test. Supplementary Fig.

S11B. P = 0.0009 (siIQGAP1), P = 0.0008 (siS1PR2), P = 0.0002 (siFXR), P = 0.0002 (siVDR)

and P = 0.0070 (siPXR) by two-tailed unpaired t-test. Supplementary Fig. S11C. P = 0.0465

(siCtrl vs siVDR) by one-way ANOVA followed by Tukey’s multiple comparison test.

Supplementary Fig. S11F. P < 0.0001 by two-tailed unpaired t-test. Supplementary Fig.

S11G. P = 0.0223 (siCtrl vs siVDR with TDCA treatment) by one-way ANOVA followed by

Tukey’s multiple comparison test. Supplementary Fig. S12C. 95% confidence interval and P

values by Cox proportional-hazards regression.

12

Fig. S1. LN metastatic tumors undergo transcriptional changes toward increased FAO. (A)

Schematic diagram for comparative transcriptomic analysis during B16F10 melanoma LN

metastasis. (B) Flow cytometry analysis of B16F10-GFP melanoma cells from primary tumor

and tumors from micro- and macro-metastatic LNs. After depleting CD45+ hematopoietic cells

by magnetic-activated cell sorting, GFP+ melanoma cells were sorted by flow cytometry for

RNA-seq. (C) Principal component analysis of the RNA-seq data from primary (PT), micro-

metastatic (Micro), and macro-metastatic (Macro) tumors (n = 4 for each group). (D)

Comparison of similarity between the tumor groups according to principal component analysis,

by pairwise inter-cluster distance. Mahalnobis distance: 4.269 [PT-Macro] vs 6.88 [PT-Micro].

13

(E and F) Comparison of top ten canonical pathways according to Ingenuity Pathway Analysis

conducted within genes annotated by gene ontology terms associated with fatty acid metabolic

process (ID: GO:0006629) between micro-metastatic tumor (E) or macro-metastatic tumor (F) at

LN, and primary tumor. P-values are calculated by Fisher’s exact t-test right-tailed. A positive or

negative z-score value indicates that a function is predicted to be increased or decreased in tumor

from metastatic LNs relative to primary tumor. (G and H) Gene set enrichment analysis of the

genes related to PPAR-α activation in micro-metastatic tumor (G) and macro-metastatic tumor

(H) at LN. FDR, false discovery rate. ****P < 0.0001.

14

Fig. S2. LN metastatic tumor exhibits increased fatty acid accumulation and oxidation. (A)

Comparison of metabolome profiles between footpad (primary tumor implantation site) and

naïve LN (n = 3 for each group). (B) Schematic diagram depicting in vivo selection and

expansion of LN metastasis-prone adapted B16F10 cells from the parental B16F10 cells. (C)

Gross appearance of metastatic pLNs, and comparisons of primary tumor volume and metastatic

tumor area between implanted parental or metastasis-adapted B16F10 cells (n = 8 for each

group). Adapted or parental cells were implanted to footpad of 8-week-old mice and pLNs were

harvested at 3rd week. Note that adapted cells form smaller primary tumor but show markedly

higher metastatic growth in the tumor-draining LNs. Scale bar, 5mm. (D) Comparison of cell

viability between the adapted cells treated with indicated concentrations of etomoxir, grown in

high glucose media, low glucose media, and low glucose media with fatty acid (palmitate 0.2

mM) (n = 5 for each group). (E) Comparison of oxygen consumption rate (OCR) for FAO

pathway between the adapted cells grown in basal media or basal media supplemented with fatty

acid (palmitate 0.2 mM) (n = 4 for each group). *P < 0.05, and **P < 0.01. Error bars are

presented as mean ± s.e.m.

15

Fig. S3. LN metastatic tumor exhibits increased fatty acid oxidation. (A) Schematic diagram

of ex vivo FAO assay with 14C-palmitic acid. (B) Comparison of production of 14C-acid soluble

metabolites between primary tumor (PT), lung metastatic tumor (Lung-MT), LN metastatic

tumor (LMT), and brown adipose tissue (BAT) by ex vivo FAO assay with 14C-palmitic acid (n =

4 for each group). (C) Comparison of 14CO2 production between PT, naïve LN, and LMT by ex

vivo FAO assay with 14C-palmitic acid (n = 3 for each). (D and E) Comparison of primary

B16F10 melanoma volume (D) and body weight (E) between PBS- and etomoxir-treated groups

(n = 12 for each group). Treatment schedule of etomoxir on B16F10 melanoma footpad

implantation model is shown in Figure 2D. *P < 0.05, **P < 0.01 and ****P < 0.0001. Error

bars are presented as mean ± s.e.m.

16

Fig. S4. Inhibition of FAO by etomoxir suppresses LN metastasis of melanoma and breast

cancer. (A) Schematic diagram depicting implantation of B16F10 cells into footpad of mice.

The LNs and lungs were harvested at day 25 after PBS or etomoxir treatment. (B to D)

Representative images of metastatic lungs (scale bar, 5 mm) and lung sections stained with H&E

(scale bars, 100 μm), and comparison of percentage of metastatic nodule formation in lung

between PBS- and etomoxir-treated groups (n = 12 for each). Arrowheads indicate metastatic

tumor nodules. (E) Schedule of etomoxir treatment after removal of primary tumor in B16F10

melanoma footpad implantation model. (F and G) Gross appearance of metastatic LNs and

comparisons of LN metastasis rate and metastatic tumor area between PBS- and etomoxir-treated

groups (n = 6 for each). Scale bar, 5 mm. (H) Schedule of etomoxir treatment to MMTV-PyMT

spontaneous breast cancer model. (I and J) Representative IF images of metastatic axillary LNs

and comparisons of metastasis rate and metastatic tumor area between PBS- and etomoxir-

treated groups (n = 8 for each). Scale bar, 500 μm. (K) Schedule of etomoxir treatment to

orthotopic 4T1 breast cancer model. (L and M) Representative IF images of metastatic axillary

LNs, and comparisons of metastasis rate and metastatic tumor area between PBS- and etomoxir-

treated groups (n = 8 for each). White dotted line demarcates metastatic tumor. Scale bar, 500

μm. *P < 0.05, and **P < 0.01. Error bars are presented as mean ± s.e.m.

17

Fig. S5. Inhibition of FAO by etomoxir suppresses tumor growth in LN. (A) Schematic

diagram depicting generation of a direct implantation of B16F10 cells into LN model. After

implantation of B16F10 cells into inguinal LNs, PBS or etomoxir (40 mg/kg) was treated daily

for 3 days before harvesting the LNs. (B and C) Gross appearance of metastatic LNs, and

comparisons of LN metastatic foci formation rate and metastatic tumor area between PBS- and

etomoxir-treated groups (n = 8 for each). Scale bar, 5 mm. (D) Schematic diagrams depicting

direct implantations of 4T1 cells into iLNs. After implantation of 4T1 cells into inguinal LNs,

PBS or etomoxir (40 mg/kg) was treated daily for 3 days before harvesting the LNs. (E)

Representative IF images of iLNs and comparison of metastatic tumor area between PBS- and

etomoxir-treated groups (n = 8 for each). Scale bar, 500 μm. (F) Schematic diagram depicting

the generation of a blood-borne lung metastasis model by injection of B16F10 cells via tail vein.

After the injection, PBS or etomoxir (40 mg/kg) was administered daily for one week before

harvesting the lungs. (G to I) Representative images of metastatic lungs (scale bar, 5 mm), lung

sections stained with H&E (scale bars, 100 μm), and comparison of mean metastatic lung

fraction between PBS- and etomoxir-treated groups (n = 6 for each). Arrowheads indicate

metastatic tumor nodules. (J) Schematic diagrams depicting tail vein injections for lung

metastasis of 4T1 cells. After the injection, PBS or etomoxir (40 mg/kg) was administered daily

for one week before harvesting the lungs. (K) Representative lung sections stained with H&E

and comparison of mean metastatic lung fraction between PBS- and etomoxir-treated groups (n =

6 for each). Arrowheads indicate metastatic tumor nodules. Scale bars, 100 μm. **P < 0.01, and

***P < 0.001. Error bars are presented as mean ± s.e.m.

18

Fig. S6. YAP activation mediates increased FAO and tumor growth of LN-metastasized

tumors. (A) Top 10 oncogenic signature genes by gene set enrichment analysis between primary

tumor and micro-metastatic tumor at LN in the B16F10 melanoma footpad implantation model.

(B) qRT-PCR analysis showing knockdown efficiency of each corresponding siRNA for

indicated gene in adapted B16F10 cells. EGFR mRNA was undetectable. (C) Comparison of

relative mRNA expressions of YAP target genes between parental and adapted B16F10 cells. (D)

19

Comparison of 14CO2 production between control- and hyperactive YAP (YAP-5SA)-B16F10

cells (n = 4 for each group). In vitro FAO assay was performed with 14C-palmitic acid. (E)

Schematic diagram depicting direct implantation of control- or YAP-5SA-expressing B16F10

cells into inguinal LN (iLN). Gross appearance of metastatic LNs and comparison of metastatic

tumor area between control- and YAP-5SA-expressing B16F10 cells (n = 8 for each group).

Scale bar, 5 mm. (F and G) IF images showing intra-tumoral area and invasive front of primary

tumors of B16F10 melanoma at footpad and breast cancer of MMTV-PyMT mouse. Inlets show

localizations of YAP in high magnification. Scale bars, 50 μm. (H) Quantifications of nuclear

localization of YAP in tumor cells in both tumors. B16F10 melanoma (n = 8); breast cancer of

MMTV-PyMT mice (n = 5). (I) IF images showing metastatic tumor nodules in the lungs of

B16F10 melanoma footpad implantation model and MMTV-PyMT spontaneous breast cancer

model. Note the predominantly cytoplasmic localizations of YAP in both tumors in high

magnification (inlets). Scale bars, 50 μm. NES; normalized enrichment score. *P < 0.05, ***P <

0.001, and ****P < 0.0001. Error bars are presented as mean ± s.e.m.

20

Fig. S7. Nuclear localization of YAP is only observed in tumors residing in LNs. (A) IF

images showing predominantly cytoplasmic YAP in B16F10 melanoma at the primary

implantation site of mouse ear skin. Scale bar, 20 μm. (B) IF images showing migrating B16F10

cells in the connecting lymphatic vessels between ear skin (primary implantation site of the cells)

and tumor-draining LN. Note that the migrating melanoma cells predominantly have cytosolic

YAP (a’) but some stromal cells predominantly have nuclear YAP (b’). Scale bars, 25 μm. (C) IF

images showing predominantly nuclear YAP (arrowheads) in B16F10 melanoma directly

implanted into inguinal LN at invasive front (right) but not in intra-tumoral region (left). Scale

bars, 50μm. (D) Quantification of tumor cells with nuclear YAP in invasive front and intra-

21

tumoral regions (n = 6 for each). (E) qRT-PCR analysis showing knockdown efficiency of

control or inducible YAP knockdown B16F10 cell line after treatment of doxycycline for two

passages in vitro (n = 6 for each group). (F) Comparison of primary tumor growth between

control and YAP-iKD groups (n = 7 for each group). (G) IF images showing pLNs that harbor

metastasized B16F10 melanoma (S100, green) in control and YAP-iKD groups. (H) IF images

showing pLN that harbor metastasized B16F10 melanoma in control and inducible YAP-

knockdown groups and their expressions of YAP (green). White dotted-lines demarcate

metastatic tumor. Scale bars, 100 μm. **P < 0.01 and ****P < 0.0001. Error bars are presented

as mean ± s.e.m.

22

Fig. S8. YAP activation is critical for LN metastasis. (A) Schedule of YAP depletion after

removal of primary tumor in the B16F10 melanoma footpad implantation model. Doxycycline

diet was started at the day after removal of primary tumor. (B and C) Gross appearance of

metastatic pLNs, and comparisons of LN metastasis rate and metastatic tumor area between

control and YAP-depleted groups (n = 6 for each). Scale bar, 5 mm. (D) Schematic diagram

depicting direct implantation of shCtrl- or shYAP-transfected B16F10 cells into inguinal LN

(iLN) of mice. Doxycycline was added into the media during culture of the B16F10 cells for

YAP depletion before the implantation of the cells. (E and F) Gross appearance of LNs, and

comparisons of LN metastatic foci formation rate and metastatic tumor area between the control

and the YAP-depleted B16F10 melanoma (n = 9 for each). Scale bar, 5mm. (G) Schematic

23

diagram depicting tail vein injection of control or YAP- depleted B16F10 cells for lung

metastasis. (H to J) Representative images of metastatic lungs (scale bar, 5 mm), lung sections

stained with H&E (scale bars, 100 μm), and comparison of mean metastatic lung fraction

between control and YAP-depleted groups (n = 6 for each). Arrowheads indicate metastatic

tumor nodules. (K) Schematic diagrams depicting direct implantations of siCtrl (control)- and

siYAP-transfected (YAP-depleted) 4T1 cells into iLNs. (L) IF images showing 4T1 metastasis in

LN. Inlets show localizations of YAP at high magnification. Note that YAP is localized in the

nuclei at invasive front (arrowheads). Scale bars, 50 μm. (M) Representative IF images of iLNs

and comparison of metastatic tumor area between control and YAP-depleted 4T1 cells (n = 8 for

each). Scale bar, 500 μm. (N) Schematic diagrams depicting tail vein injections for lung

metastasis of siCtrl (control)- and siYAP-transfected (YAP-depleted) 4T1 cells. (O)

Representative lung sections stained with H&E and comparison of mean metastatic lung fraction

between control and YAP-depleted 4T1 cells. (n = 6 for each). Arrowheads indicate metastatic

tumor nodules. Scale bars, 100 μm. *P < 0.05, **P < 0.01, and ***P < 0.001. Error bars are

presented as mean ± s.e.m.

24

Fig. S9. Elevated bile acids activate YAP of the tumor cells to promote growth of LN

metastatic melanoma. (A) IF images showing Ki-67+ B16F10-GFP cells (arrowheads) in LN

metastatic tumor, in invasive front (a’) and intra-tumoral area (b’). Scale bar, 100 μm. (B) IF

images showing GLUT1+ hypoxic region in the LN metastatic tumor of B16F10 melanoma.

White dotted-lines demarcate the metastatic tumor. Note that the GLUT1+ region lacks of CD31+

/PDGFRβ+ blood vessels. Scale bar, 200 μm. (C) Comparisons of relative abundances of

cholesterol among primary tumor and LN metastatic tumor in the B16F10 melanoma footpad

implantation model, and footpad tissue and naïve LN of control mice (n = 3 for each group). (D)

Comparison of relative abundances of circulating cholic acid (CA), deoxycholic acid (DCA),

25

chenodeoxycholic acid (CDCA), and lithocholic acid (LCA) in the blood and lymph from normal

control mice without tumor and LN metastatic B16F10 melanoma-bearing mice (n = 4 for each

group). (E) Gene set enrichment analysis (left) and heatmap (right) of the genes related to bile

acid metabolism between LN macro-metastatic tumor and primary tumor. (F) Arbitrary ratios of

phosphorylated YAP per total YAP in adapted B16F10 cells during indicated time after

treatment of TDCA (100 μM). See Fig. 4C for immunoblot images. (G and H) Immunoblot

analysis of phosphorylated and total YAP in adapted B16F10 cells (G) at 30 min after treatment

with taurodeoxycholic acid (TDCA), a hydrophobic bile acid, for indicated concentrations, and

(H) after treatment with taurocholic acid (TCA), a hydrophilic bile acid which cannot freely

enter the cell. (I) Schedule of TDCA treatment to B16F10 melanoma footpad implantation

model. (J to L) Gross appearance of metastatic pLN, and comparisons of primary tumor volume

and metastatic tumor area between PBS- and TDCA-treated groups (n = 8 for each). Scale bar,

5mm. AU, arbitrary unit. *P < 0.05, and ***P < 0.001. Error bars are presented as mean ± s.e.m.

26

Fig. S10. LN metastatic tumor generates bile acids from cholesterol to activate YAP of the

tumor cells. (A) Immunoblot analysis of phosphorylated and total YAP in adapted B16F10 cells

after treatment with methyl-β-cyclodextrin (MβCD, 700 M), a solubilizer for cholesterol. (B

and C) Arbitrary ratios of phosphorylated YAP per total YAP in adapted B16F10 cells during

indicated time after treatment of cholesterol (100 μM). See Fig. 4, D and E for immunoblot

images. In B, MβCD was used as a solubilizer for cholesterol. In C, adapted B16F10 cells were

transfected with two different (#1 and #2) CYP7A1 siRNA and then treated with cholesterol for

60 min. (D) Immunoblot analysis and quantification of CYP7A1 levels in the primary (PT) and

LN metastatic melanomas (LMT). (E) Thin-layer chromatography (TLC) analysis for generation

of bile acids from 14C-cholesterol in parental and adapted B16F10 cells. 14C-labeled bands

corresponding to cholesterol (Chol), 7α-hydroxycholesterol (7αOHChol), putative CDCA

(CDCA#), and putative CA (CA#) are detected in adapted cells but not in parental cells. Origin,

origin of sample spotting. Representative image of 3 independent experiments. (F) Comparisons

of 14C-cholesterol (Chol) conversion into 7αOHChol, CDCA#, and CA# between adapted

B16F10 cells transfected with control or CYP7A1 siRNA (n = 3 for each group). Origin, origin

of sample spotting. Representative image of 3 independent experiments. Relative radioactivity

27

(%) of cholesterol and each putative bile acid species per total radioactivity are shown. (G)

Schematic diagram depicting direct implantations of siCtrl (control)- or siCYP7A1-transfected

(CYP7A1-depleted) B16F10 cells into inguinal LNs. (H and I) Gross appearance of LNs and

comparison of metastatic tumor areas between control and CYP7A1-depleted B16F10 melanoma

(n = 8 for each). Scale bar, 5 mm. (J) IF images showing control and CYP7A1-depleted B16F10

cells in LNs. Inlets show localization patterns of YAP at high magnification. Scale bars, 100 μm.

(K) Quantification of nuclear YAP localization in B16F10 cells of control and CYP7A1-depleted

B16F10 tumor groups (n = 5 for each). AU, arbitrary unit. *P < 0.05, **P < 0.01, ***P < 0.001

and ****P < 0.0001. Error bars are presented as mean ± s.e.m.

28

Fig. S11. Bile acids promote LN metastasis mainly via VDR. (A) Comparison of relative

mRNA levels of IQGAP1, S1PR2, FXR, VDR, and PXR in the liver and skeletal muscle of

normal healthy mice and the LN metastatic tumor. Relative 2-∆CT value is shown (n = 3 for each

group). (B) Knockdown efficiency of each corresponding siRNA for the indicated gene by qRT-

PCR. (C) Gross appearance of LNs and comparisons of metastatic tumor area between direct

implantation of control or IQGAP1-, S1PR2-, FXR-, VDR- or PXR-depleted B16F10 cells at 3

days after direct implantation into inguinal LNs (n = 6 for each). Scale bar, 5 mm. (D) IF images

showing localization patterns of YAP in the control or IQGAP1-, S1PR2-, FXR-, VDR- or PXR-

depleted B16F10 cells of LN metastatic tumors. Inlets, highly magnified view of each square

zone. Scale bars, 100 μm. (E and F) Gross appearance of LNs and comparison of metastatic

tumor area between control and VDR-depleted B16F10 tumors implanted directly in LNs of

mice (n = 8 for each group). Scale bar, 5 mm. (G) Immunoblot analyses (top) and quantifications

(bottom) of phosphorylated and total YAP in the adapted B16F10 cells transfected with control

or VDR siRNA with or without treatment of TDCA (100 μM) for 30 min. *P < 0.05, **P < 0.01,

***P < 0.001, and ****P < 0.0001. Error bars are presented as mean ± s.e.m.

29

Fig. S12. Nuclear localization of YAP in the metastatic tumor cells at invasive front of

tumor draining LN is correlated with worse prognosis in patients with melanoma. (A)

Immunohistochemical (IHC) image showing strong YAP expression in the metastatic nodule of

tumor draining LN in patients of melanoma. Scale bar, 2 mm. (B) IHC image showing

predominantly cytoplasmic YAP in primary tumor of patients with melanoma. Scale bar, 100

μm. (C) Multivariate analysis of distant metastasis-free survival conducted with age, primary

tumor depth, and status of YAP activation at invasive front. HR, hazard ratio; CI, confidence

interval; P, p-value.

30

Fig. S13. Schematic diagram summarizing the proposed model for the mechanism of tumor

LN metastasis. Tumor cells undergoing metastasis to tumor-draining LNs adapt to the LN

microenvironment through YAP-dependent adjustment to fatty acid oxidation and usage of the

plentiful fatty acids within the microenvironment as fuel. Highly accumulated bile acids as

potential molecular triggers of YAP activation mainly through vitamin D receptor (VDR) in

metastatic LNs.

31

Table S1. Clinicopathologic characteristics of the melanoma patients.

Patient

ID

Localization of YAP

at invasive front Sex Age

Depth of invasion (mm) Stage†

Previous therapy

P1 Nuclear Male 50 10 IIIC None

P2 Nuclear Female 46 8 IIIC None

P3 Nuclear Male 71 3 IIB None

P4 Nuclear Female 79 2.2 IIIB None

P5 Nuclear Male 56 7 IIIC None

P6 Nuclear Male 74 2.2 IIIB None

P7 Nuclear Male 52 30 IV None

P8 Nuclear Male 56 9.2 IIIC None

P9 Cytoplasmic Female 49 8 IIIC None

P10 Cytoplasmic Male 67 5 IIIC None

P11 Cytoplasmic Female 73 3.5 IV None

P12 Cytoplasmic Male 61 4 IIIB None

P13 Cytoplasmic Female 81 1.1 IIIC None

P14 Cytoplasmic Male 62 8 IIIC None

P15 Cytoplasmic Female 58 6 IIIC None

P16 Cytoplasmic Male 82 1.5 IIIB None

P17 Cytoplasmic Male 65 8 IIIB None

P18 Cytoplasmic Male 58 1.6 IIIA None

P19 Cytoplasmic Male 52 2 IV None

P20 Cytoplasmic Female 69 6 IV None

P21 Cytoplasmic Female 68 8 IV None

†AJCC stage at the time of sentinel LN removal.