Supporting InformationDavare et al. 10.1073/pnas.1319583110SI MethodsIn Vivo Inhibitor Treatment of Tumor-Bearing Mice. All animal ex-periments were performed in accordance with a protocol ap-proved by the Memorial Sloan-Kettering Institutional AnimalCare and Use Committee. Foretinib was dissolved in DMSO,aliquoted, and stored at −80 °C. Before use, aliquots were fur-ther diluted in 1% hydroxypropylmethylcellulose/0.2% SDS. Cri-zotinib was reconstituted in 1% hydroxypropylmethylcellulose/0.2% SDS, aliquoted, and stored at −80 °C. Tumor cell lines ex-pressing either the FIG–ROS kinase fusion or shRNA againstmurine phosphatase and tensin homolog (Pten) were s.c. injectedinto 11-wk-old female immunodeficient Crl:Nu-Foxn1nu miceweighing 26 g. Upon reaching a tumor diameter of 4–6 mm, micewere treated with foretinib (25 mg/kg; molecular weight = 632.65),crizotinib (25 mg/kg; molecular weight = 450.34), or vehicle con-trol by oral gavage once daily for 9 consecutive days. Calipermeasurements were taken at the time of treatment initiation (T0)and at 24 h after administration of the last dose (Tend). Data in thewaterfall plots are calculated as follows: ((tumor volume Tend/tumor volume T0) − 1). Tumor volume is calculated as: 0.5 × L ×W2 with L > W.
Isolation of Primary Liver Progenitor Cells and Tumor Cell LineGeneration. A detailed description of the production and char-acterization of cholangiocarcinomas from these cells are de-scribed in the companion paper in PNAS (1). Isolation of liverprogenitor cells was performed as previously described (2).Briefly, to produce tumors, hepatoblasts of the genotypeAlbCre+/−; lsl KrasG12D+/−; p53R172H/loxp were isolated fromembryonic day (ED) 14.5 mouse embryos and retrovirallytransduced with either the FIG–ROS fusion or a potent shorthairpin RNA against Pten (shPten.1522). At 48 h post trans-duction, 1 × 106 cells were resuspended in 25 μL Matrigel (BD)and injected subcapsular into the livers of immunodeficient (Crl:Nu-Foxn1nu) nude mice. Upon tumor formation, tumors wereexplanted and subjected to collagenase digestion (Sigma;C1538). Crude digests were plated onto gelatin-coated plates.Enrichment of tumor cells was enhanced by differential trypsi-nization or cell sorting. These murine cholangiocarcinoma tu-mor-derived cell lines expressing FIG–ROS (lines 3 and 4) andshPten (lines 1 and 2) were further cultured and maintained inDMEM-high glucose medium (Invitrogen) supplemented with10% FCS (Invitrogen), L-glutamine (Invitrogen), and penicillin/streptomycin (Invitrogen).
Plasmid Construction. The FIG–ROS–S fusion gene was synthe-sized using the GeneArt service (Invitrogen). FIG–ROS–S isdenoted as FIG–ROS here. FIG–ROS was further subclonedinto the retroviral vector, pMSCV–IRES–GFP (pMIG) using theGateway Cloning system (Invitrogen). SLC–ROS–S (SLC–ROS)
was cloned from cDNA made from the non-small cell lungcancer cell line, HCC78. Briefly, using sense primers that wereSLC34A2 N-terminus specific (5′ CAC CAT GGC TCC CTGGCC TGA ATT GG) and anti-sense primers that are ROS1-specific (5′ TTA ATC AGA CCC ATC TCC ATA TCC ACTGTG AGT G), we were able to amplify both SLC–ROS–L andSLC–ROS–S from HCC78 cDNA. We individually cloned SLC–ROS–L and SLC–ROS–S after gel extraction of the PCRproducts into the Gateway Cloning system compatible entryvector pENTR–D/TOPO (Invitrogen) and further subclonedinto the pMIG retroviral vector as described above. The FIG–
ROS point mutations were created using the Quikchange site-directed mutagenesis kit (Agilent Technologies) according to themanufacturer’s protocol.
IL-3 Withdrawal/Transformation Assays. Parental Ba/F3, pMIG-alone, FIG–ROS (wild-type or ROS1 KD mutant variants) orSLC–ROS-expressing Ba/F3 cells (3 × 106 cells total) werewashed three times with 50 mL of RPMI-10% FBS media andresuspended in a new 6 mL of RPMI-10% FBS media. The totalnumber of viable cells was counted every other day using GuavaViaCount reagent and a Guava Personal Cell Analysis flow cy-tometer (Guava Technologies). If cells grew to cell densities>1.5 × 106/mL in withdrawal media, the cells were centrifugedand resuspended in fresh media to keep a final culture density of0.5 × 106/mL.
Immunoblotting. Ba/F3 FIG–ROS, SLC–ROS, HCC78, and mu-rine cholangiocarcinoma cell lines were treated with the in-dicated concentrations of inhibitors for 1–2 h. In the case of Ba/F3cells, after treatment 5 × 106 cells were pelleted, washed once inice-cold PBS, and lysed in 200 μL of cell lysis buffer (Cell Sig-naling Technology) supplemented with 0.25% deoxycholate,0.05% SDS, and protease and phosphatase inhibitors. Afterprotein quantification, equal amounts of protein containing lysatewere either used for immunoprecipitation where indicated orextracted with SDS sample buffer for 15 min at 80 °C. Proteinswere transferred to Immobilon-FL membranes (Millipore) andsubjected to immunoblot analysis with antibodies specific forphospho-ROS1 [Cell Signaling Technology (CST); 3078], totalROS1 (CST; 3266), phospho-SHP2 (CST; 3751), total SHP2(CST; 3752), phospho-STAT3 (CST; 9145), total STAT3 (CST;4904), phospho-ERK1/2 (CST; 9101), total ERK2 (Santa Cruz;sc-1647), phospho-S6 (CST; 4858), total S6 (CST; 2216), phos-pho-SRC (CST; 2105), total SRC (CST; 2110), α-tubulin (Sigma;T6199), and β-actin (Sigma; A1978). We used the LI-COR Od-yssey imaging system or the Bio-Rad Chemidoc imaging stationand followed the manufacturer’s protocol for immunoblot de-tection with use of IR dye or HRP-conjugated secondary anti-bodies, respectively.
1. Saborowski A, et al. (2013) Mouse model of intrahepatic cholangiocarcinoma validatesFIG–ROS as a potent fusion oncogene and therapeutic target. Proc Natl Acad Sci USA,10.1073/pnas.1311707110.
2. Zender L, et al. (2005) Generation and analysis of genetically defined liver carcinomasderived from bipotential liver progenitors. Cold Spring Harb Symp Quant Biol 70:251–261.
Davare et al. www.pnas.org/cgi/content/short/1319583110 1 of 8
A. B.
100 500 1,000 2,000 3,0000
1
2
3
Apop
tosis
Indu
ctio
n(F
old
over
bas
al)
Kinase Inhibitor (Nanomolar)
Crizotinib
Foretinib
**
*****
*HCC78
**
Ba/F3: FIG-ROS-S
Ba/F3: SLC-ROS-S
Apop
tosis
Indu
ction
(Fold
ove
r bas
al)
0
10
20
30
0
5
10
15
20
0
10
20
30
40
Apop
tosis
Indu
ction
(Fold
ove
r bas
al)Ap
opto
sis In
ducti
on(F
old o
ver b
asal)
Ba/F3: ALK F1174L
0.09
0.3
0.4 0.5 0.25
15 nM Crizotinib
150 nM Crizotinib
15 nM Foretinib
15 nM NVP-TAE684
15 nM Gö6976
Inhibitor Treatment
Ba/F3 ALK F1174L
Vehicle
0
20
40
60
80
100
% A
popt
osis
(Ann
exin
+ve
)
Ba/F3 EML4-ALK
Vehicl
e 5 25 250
500
1000
Gö6976
15nM
TA
E684
0
20
40
60
Inhibitor Treatment (nM)
% A
popt
osis
(Ann
exin
+ve
)
5 25 250
50010
00
Foretinib
15nM
TA
E6845 20 40 20
075
0
Gö6976
Inhibitor Treatment (nM)
Foretinib
5 20 40 200
750
C.
D.
Fig. S1. Induction of apoptosis with inhibitor treatment in Ba/F3 and HCC78 cells. (A) Apoptosis induction in Ba/F3 FIG–ROS, SLC–ROS, and ALKF1174L cells after48 h of 15-nM inhibitor treatment as indicated. Number of cells that were Annexin V and 7-Aminoactinomycin D positive after inhibitor treatment werenormalized to number of cells staining positive with these markers with vehicle treatment (basal). A minimum of 2,000 cells were counted per condition. Rateof apoptosis is shown as fold over basal. Basal apoptosis in Ba/F3 FIG–ROS, SLC–ROS, and ALK F1174 cells was 3.6%, 2.8%, and 1.2%, respectively. (B and C)Apoptosis induction (Annexin V and 7-AAD staining) in Ba/F3 ALKF1174L and EML4–ALK after 48 h of inhibitor treatment as indicated in the graphs. (D) Ap-optosis induction in HCC78 cells after treatment with foretinib and crizotinib for 72 h. Method of detection and analysis was identical to Ba/F3 cells. Basal(vehicle treated) apoptosis in HCC78 cells was 11% of total cell population. *P < 0.05, **P < 0.01 and ***P < 0.0001 by t test.
Davare et al. www.pnas.org/cgi/content/short/1319583110 2 of 8
A.
B.
pROS1
tROS1
Tubulin .01 0.1 1 .01 0.1 1 .01 0.1 1
Crizotinib Foretinib .01 0.1 1
NIH3T3 FIG-ROS NIH3T3 SLC-ROS
Crizotinib Foretinib
Concentration (nM): 2.5 25 250 2.5 25 250
Crizotinib
Foretinib
FIG-ROS SLC-ROS
0
5
10
15
Aver
age
Col
ony
#
NIH3T3 FIG-ROS
0
5
10
15
20
25
Vehicle
Crizotinib
Foretinib
Veh. 2.5 25 250 2.5 25 250
Crizotinib Foretinib
Ave
rage
Col
ony
#
Veh.
Crizotinib ForetinibNIH3T3 SLC-ROS
2.5 25 250 2.5 25 250(nM)(nM)
Veh.
Veh.
Vehicle
Fig. S2. Foretinib and crizotinib suppress ROS1 fusion phosphorylation and diminish anchorage-independent colony formation in transformed NIH 3T3 cells.(A) Immunoblot analysis of ROS1 fusion phosphorylation from NIH 3T3 FIG–ROS and SLC–ROS cells treated with varying concentrations of crizotinib andforetinib for 1 h. (B, Upper) Representative images of NIH 3T3 FIG–ROS and SLC–ROS colony formation in soft agar with and without varying dose of crizotiniband foretinib. (Lower) Quantification of colony number formed from vehicle or inhibitor-treated NIH 3T3 FIG–ROS and SLC–ROS cells. Graph shows averagecolony number ± SEM from four independent wells.
Davare et al. www.pnas.org/cgi/content/short/1319583110 3 of 8
A.
B.
C.
Crizoti
nib
Foretin
ib
MGCD-265
SGX-523
JNJ-3
8877
605
100
1000
10000
100000
Kinase Inhibitor
HCC78 (SLC-ROS)
(Nan
omol
ar)
IC50
10 minutes 18 hours
Vehicle
Crizotinib50nM
Foretinib50nM
0%
25%
50%
75%
100%
% G
ap C
losu
re
Vehicl
e 0.05 0.5 1Crizotinib Foretinib
HCC78
Crizoti
nib
Foretin
ib
Erlotin
ib100
1000
10000 HCC78 (SLC-ROS)
KARPAS 299 (NPM-ALK)
SUDHL1 (NPM-ALK)
PC9 (EGFR L1848R)
HCC4011 (EGFR L1848R)
Kinase Inhibitor
Con
cent
ratio
n(N
anom
olar
)
Fig. S3. Analysis of inhibitor effect on SLC-ROS expressing non-small cell lung cancer cell line HCC78 cell migration and assessment of foretinib specificity inHCC78 cells. (A, Left) Representative images show gap at 10 min and gap closure at 18 h post “scratch” after vehicle, 50 nM crizotinib, or 50 nM foretinibtreatment. (Right) Quantification of gap closure determined by measuring micrometer length of a straight line drawn across the gap from imaging the samefield of view at initiation of treatment and at 18 h (end of treatment). (Scale bar, 200 μm.) (B) IC50 values for HCC78, KARPAS 299 (lymphoma), SUDHL1(lymphoma), PC9 (lung adenocarcinoma), and HCC4011 (lung adenocarcinoma) cells treated with crizotinib, foretinib, and erlotinib as determined fromnonlinear regression curve fit analysis of the dose–response curves. (C) IC50 values for HCC78, treated with MET kinase inhibitors crizotinib, foretinib, MGCD-265, SGX-523, and JNJ-38877605 as determined from nonlinear regression curve fit analysis of the dose–response curves.
Davare et al. www.pnas.org/cgi/content/short/1319583110 4 of 8
Line 1shPTEN
Line 2shPTEN
Line 3FIG-ROS
Line 4FIG-ROS
1 ve
hicl
e
1 fo
retin
ib 2
5mg/
kg
1 cr
izot
inib
25m
g/kg
2 ve
hicl
e
2 fo
retin
ib 2
5mg/
kg
2 cr
izot
inib
25m
g/kg
3 ve
hicl
e
3 fo
retin
ib 2
5mg/
kg
3 cr
izot
inib
25m
g/kg
4 ve
hicl
e
4 fo
retin
ib 2
5mg/
kg
4 cr
izot
inib
25m
g/kg
0
500
1,000
p = 0.068
p = 0.058
p < 0.0001
p <0 .0001Tu
mor
wei
ght (
mg)
A.
B.
Vehicle Foretinib Crizotinib
Line
2, s
hPte
nLi
ne 4
, FIG
-RO
S
Fig. S4. Inhibitor-treated FIG–ROS and shPten tumor weight and histopathology images. (A) Scatter plot shows the individual weight distribution of treatedtumors from lines 1 and 2 for shPten tumors and lines 3 and 4 for FIG–ROS-expressing tumors. Statistical significance determined with paired t test. (B)Representative sections of H&E stained tumors from line 2 (shPten) and line 4 (FIG–ROS) after 9 d of inhibitor treatment as indicated. (Scale bar, 500 μm.)
Davare et al. www.pnas.org/cgi/content/short/1319583110 5 of 8
0.E+001.E+082.E+083.E+084.E+085.E+086.E+087.E+088.E+089.E+081.E+09
0 2 4 6 9
Tota
l Via
ble
Cel
ls
Days post IL-3 withdrawal
BaF3 ParentalC2060GV2098IG1971EL1982FL1947R E1935Gwildtype
A.
0
20
40
60
80
100
120
140 siNTsiROS1
E1935
G
L194
7R
C2060
G
L198
2F
G1971
E
V2098
I
Wild
type
% C
ell V
iabi
lilty
(nor
mal
ized
to m
ock
trans
fect
ed)
B.
L1982
G1971
L1947
V2098
C2060
ActivationLoop
αCNucleotide binding Loop
ROS1 kinase domain homology model
500
750
1000
1500
2000
2500
0
10
20
30
40
50
Crizotinib (nM)
% R
esis
tant
Clo
nes
(of t
otal
wel
ls p
late
d)
C.
E.
D.
Crizotinib (nM) Total # of wells500 192750 1921000 2881500 3842000 4802500 480
Fig. S5. Recovered Ba/F3 FIG–ROS crizotinib-resistant clones remain dependent on FIG–ROS and are independently transforming in Ba/F3 cells. (A) Table showsthe number of wells of 96-well plates surveyed with 500; 750; 1,000; 1,500; 2,000; and 2,500 nM crizotinib after ENU-induced mutagenesis. (B) Number of wellswith crizotinib-resistant clones for each concentration that were recovered, expressed as a percentage of the total number of wells seeded. (C) Viability of Ba/F3 FIG–ROS mutant cell lines recovered from ENU-mutagenesis screen after electroporation with siRNA for ROS1 kinase domain (siROS1) and nontargeting(siNT). Viability of siNT and siROS1-transfected cells was normalized to mock (sterile water)-transfected cells. (D) Homology model of ROS1 kinase domainbound to crizotinib (magenta) and foretinib (orange). The ligands are shown in colored stick representation, whereas the protein is shown in gray ribbon.Differences in the activation loop conformations are highlighted by coloring according to the respective ligands. Residues that were found to confer resistanceto crizotinib when mutated are highlighted in red and shown as stick representations. (E) IL-3 withdrawal assay for Ba/F3 cells transfected with wild-type orindicated mutant FIG–ROS. Total viable cell number was determined by counting cells on days 2, 4, 6, and 9 after IL-3 withdrawal.
Davare et al. www.pnas.org/cgi/content/short/1319583110 6 of 8
10-4 10-3 10-2 10-1 100 1010.00
0.25
0.50
0.75
1.00
1.25FIG-ROS C2060G
FIG-ROS V2098I
FIG-ROS G1971E
FIG-ROS L1982F
FIG-ROS L1947R
FIG-ROS E1935G
FIG-ROS Wildtype
Crizotinib
Cel
l V
iabi
lity
(Nor
mal
ized
to V
ehic
le)
10-4 10-3 10-2 10-1 100 1010.00
0.25
0.50
0.75
1.00
1.25
Foretinib
Cel
l V
iabi
lity
(Nor
mal
ized
to V
ehic
le) FIG-ROS C2060G
FIG-ROSV2098I
FIG-ROSG1971E
FIG-ROS L1982F
FIG-ROSL1947R
FIG-ROSE1935G
FIG-ROSWildtype
B.
C.
A.
Vehi
cle5n
M50
nM50
0nM
5nM
50nM
50
0nM
Tubulin
WT
V2098I
L1982F
C2060G
L1947R
G1971E
Crizotinib Foretinib
pROS1
tROS1
Tubulin
pROS1tROS1
Tubulin
pROS1tROS1
Tubulin
pROS1tROS1
Tubulin
pROS1tROS1
Tubulin
pROS1tROS1
Ba/F3 FIG-ROS Crizotinib IC50 (nM) Foretinib IC50 (nM)Wildtype 37.5 2.03C2060G 690 13.6V2098I 901 9.9G1971E 605 8.69L1982F 2450 15.27L1947R 1420 17.87E1935G 350 6.6
D.
Fig. S6. Crizotinib and foretinib sensitivity of Ba/F3 FIG–ROS wild-type and kinase domain mutant-expressing transformed cells. (A and B) Dose–response curveshowing growth of Ba/F3 FIG–ROS wild-type and kinase domain mutant cells after 72 h exposure with varying concentrations of crizotinib (A) and foretinib (B)as normalized to vehicle-treated cells. The values are means ± SEM from three independent experiments with triplicate wells. (C) Table showing IC50 values ofcrizotinib and foretinib for Ba/F3 FIG–ROS kinase domain mutants. (D) Immunoblot analysis of wild-type and kinase domain mutant FIG–ROS phosphorylationat varying doses of crizotinib and foretinib in Ba/F3 cells. Cropped images representative of two independent experiments are shown.
Davare et al. www.pnas.org/cgi/content/short/1319583110 7 of 8
Other Supporting Information Files
Dataset S1. List of inhibitors, 384-well plate layout, and the normalized FIG–ROS cell viability data for all of the inhibitors tested in thehigh-throughput inhibitor platform
Dataset S1
Dataset S2. In vitro binding affinities (Kd) of a subset of kinase inhibitors for ROS1, ALK, MET, EGFR, PDGFRA, and/or IGF1R aspreviously reported
Dataset S2
Davare et al. www.pnas.org/cgi/content/short/1319583110 8 of 8