ORIGINAL ARTICLE

Targeting FGF19 inhibits tumor growth in colon cancer xenograft and

FGF19 transgenic hepatocellular carcinoma models

LR Desnoyers1, R Pai2, RE Ferrando2, K Hotzel2, T Le3, J Ross3, R Carano3, A D’Souza1,J Qing1, I Mohtashemi4, A Ashkenazi1 and DM French2

1Department of Molecular Oncology, Genentech Inc., South San Francisco, CA, USA; 2Department of Pathology, Genentech Inc.,South San Francisco, CA, USA; 3Department of Biomedical Imaging, Genentech Inc., South San Francisco, CA, USA and4Department of Protein Chemistry, Genentech Inc., South San Francisco, CA, USA

Although fibroblast growth factor 19 (FGF19) canpromote liver carcinogenesis in mice its involvement inhuman cancer is not well characterized. Here we reportthat FGF19 and its cognate receptor FGF receptor 4(FGFR4) are coexpressed in primary human liver, lungand colon tumors and in a subset of human colon cancercell lines. To test the importance of FGF19 for tumorgrowth, we developed an anti-FGF19 monoclonal antibodythat selectively blocks the interaction of FGF19 withFGFR4. This antibody abolished FGF19-mediated activ-ity in vitro and inhibited growth of colon tumor xenograftsin vivo and effectively prevented hepatocellular carcino-mas in FGF19 transgenic mice. The efficacy of theantibody in these models was linked to inhibition ofFGF19-dependent activation of FGFR4, FRS2, ERK andb-catenin. These findings suggest that the inactivation ofFGF19 could be beneficial for the treatment of coloncancer, liver cancer and other malignancies involvinginteraction of FGF19 and FGFR4.Oncogene (2008) 27, 85–97; doi:10.1038/sj.onc.1210623;published online 25 June 2007

Keywords: fibroblast growth factor; hepatocellularcarcinoma; FGF19; colon adenocarcinoma; FGFR4;b-catenin

Introduction

The fibroblast growth factor (FGF) family is composedof 22 structurally related polypeptides that bind to fourreceptor tyrosine kinases (FGFR1–4) and one kinase-deficient receptor (FGFR5) (Ornitz and Itoh, 2001;Sleeman et al., 2001; Eswarakumar et al., 2005). FGFinteractions with FGFR1–4 result in receptor homo-dimerization and autophosphorylation, recruitmentof cytosolic adaptors such as fibroblast growth factorreceptor substrate 2 (FRS2) and initiation of multiple

signaling pathways (Powers et al., 2000). FGFs andFGFRs play important roles in development and tissuerepair by regulating cell proliferation, migration, chemo-taxis, differentiation, morphogenesis and angiogenesis.Amplification or overexpression of fgf2–4 and 8 is

associated with the pathogenesis of leukemias andsarcomas as well as prostate, stomach, pancreas, bladderand colon cancers (Liscia et al., 1989; Huang et al., 1993;Dorkin et al., 1999; Marsh et al., 1999; Kiuru-Kuhlefeltet al., 2000; Krejci et al., 2001; Mattila et al., 2001;Ruohola et al., 2001; Shimokawa et al., 2003; Zaharievaet al., 2003; Holzmann et al., 2004). In addition, fgfr1–4overexpression, translocation, polymorphisms and trun-cations are associated with neoplastic transformation inmyeloma and breast, stomach, colon, bladder andcervical cancer (Jaakkola et al., 1993; Chesi et al.,1997, 2001; Richelda et al., 1997; Xiao et al., 1998;Cappellen et al., 1999; Popovici et al., 1999; Jang et al.,2000, 2001; Bange et al., 2002; Jeffers et al., 2002;Gowardhan et al., 2005).We previously showed that FGF19 ectopic expression

in mice promotes hepatocyte proliferation, hepatocel-lular dysplasia and neoplasia (Nicholes et al., 2002). Toinvestigate FGF19 association with human cancer, weevaluated its expression in primary tumors and testedthe consequence of therapeutic FGF19 neutralization inrelevant tumor models. We show here that FGF19 andits cognate receptor FGFR4 are coexpressed in primaryhuman liver, lung and colon tumors. We evaluatedFGF19 binding to FGFR5 and developed an antibodyto neutralize this interaction. Administration of thisantibody inhibited tumor growth in two human coloncancer models and FGF19 transgenic hepatocellularcarcinomas (HCCs). These findings suggest that thetherapeutic inactivation of FGF19 could be beneficialfor the treatment of colon and liver cancer.

Results

FGF19 and FGFR4 are expressed in human tumorsFgf19 was overexpressed in 6 out of 10 colon adenocarci-nomas (Figure 1a) and in 7 out of 10 lung squamous cellcarcinomas (SCC) relative to normal tissues (Figure 1b).

Received 10 January 2007; revised 20 April 2007; accepted 15 May 2007;published online 25 June 2007

Correspondence: Dr DM French, Genentech Inc., MS 72B, 1 DNAWay, South San Francisco, CA 94080, USA.E-mail: dfrench@gene.com

Oncogene (2008) 27, 85–97& 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00

www.nature.com/onc

Compared to normal tissues, fgfr4 expression was notsignificantly altered in colon tumors but appeared down-regulated in SCC tumors (Figures 1a and b).

To localize fgf19 and fgfr4 expression in tumortissues, we performed in situ hybridization. MessengerRNA for both genes was prominent in neoplastic

FGF19 ISH

FGFR4 ISH

Colon Adenocarcinoma

Hepatocellular Carcinoma

Cirrhotic Liver

FGF19 ISH

Lung Squamous Cell Carcinoma

FGF19 IHC

FGF19 FGFR4

Colon Adenocarcinoma

0

10

100

1,000

10,000 FGF19 FGFR4

Tumor Normal

Tumor Normal

Rel

ativ

e E

xpre

ssio

n

0

10

100

1,000

10,000

Rel

ativ

e E

xpre

ssio

n

Lung Squamous Cell Carcinoma

FGFR4 ISH

FGFR4 ISH

FGF19 ISH

Tissue Samples (tumor and normal)

Tissue Samples (tumor and matched normal)

FGF19 ISH

FGF19 IHC

FGF19 IHC

FGF19 IHC

a d

e

f

c

b

FGF19 and carcinomas in humansLR Desnoyers et al

86

Oncogene

epithelial cells in colon adenocarcinomas and lung SCC(Figures 1c and d). In a tissue microarray comprising 35colon adenocarcinoma cases, 26 (74%) had positivesignal for fgf19 messenger RNA (mRNA) and 27 (77%)had positive signal for fgfr4 mRNA. Of 14 lung SCCcases, 14 (100%) had weak to moderate positive signalfor fgf19 mRNA and 13 (93%) had weak positive signalfor fgfr4 mRNA. In addition, neoplastic epithelial cellsshowed strong FGF19 staining by immunohistochem-istry in both colon adenocarcinomas (Figure 1c) andlung SCC (Figure 1d).Because systemic FGF19 expression in transgenic

mice promotes HCC (Nicholes et al., 2002), we alsoevaluated fgf19 and fgfr4 mRNA expression in liversamples. Of 50 cases of HCC, 23 (46%) demonstratedpositive signal for fgf19 and 30 (60%) for fgfr4. Bothgenes were expressed in the neoplastic hepatocytes(Figure 1e). The neoplastic hepatocytes also showedstrong FGF19 staining by immunohistochemistry(Figure 1e). These results demonstrate that FGF19and FGFR4 are expressed in several types of humancancers.Because cirrhosis often precedes HCC, we evaluated

FGF19 expression in cirrhotic liver. These samplesshowed strong FGF19 mRNA and protein signals inregenerative nodule hepatocytes (Figure 1f), suggestingthat FGF19 expression occurs early during liverneoplastic progression.

FGF19 and FGFR4 are expressed in human tumor celllines and xenograft tissuesTo assess further FGF19 and FGFR4 association withhuman cancer and to identify potential preclinicalmodels, we evaluated mRNA and protein expressionin a panel of colon, breast and liver tumor cell lines. Wefound fgf19 expression in a subset of colon cancer celllines, including Colo205, Colo201, SW620, HT29,HCT116 and SW480 (Figure 2a). SNU185, SNU398,MCF7 and all colon cancer cell lines tested expressedfgfr4. With the exception of HT29, which did notexpress FGF19 protein, the FGF19 and FGFR4 proteinlevels agreed with their mRNA expression in these celllines (Figure 2b). The electrophoretic mobility of thesecreted FGF19 was consistent with its expectedmolecular mass. However, we also detected severaladditional lower molecular mass bands possibly repre-senting potential truncated protein.To verify that FGF19 expression is maintained

in vivo, we immunostained colon cancer cell line-derivedtumor xenografts (Figure 2c). Colo205 and Colo201xenografts showed strong FGF19 staining in most

Figure 1 FGF19 and FGFR4 are expressed in cancer. Fgf19 and fgfr4 mRNA expression in normal and tumor samples from colon(a) and lung (b) evaluated by semi quantitative reverse transcription (RT)–PCR (normal tissues are indicated with an ‘N’ and tumortissues are indicated with a ‘T’). Samples evaluated for fgf19 expression are on the left and the same samples evaluated for fgfr4expression are on the right. For (a) the normal and tumor tissue are from the same patient; for (b) the normal and tumor tissue arefrom different patients. Bright-field (top left panels) and dark-field (top right panels) illumination of in situ hybridization with fgf19 orfgfr4 riboprobes (� 100 magnification for colon and cirrhotic liver; � 200 magnification for Lung and HCC) and FGF19 IHC (lowerleft panel � 100 magnification and lower right panel � 600 magnification) in colon adenocarcinoma (c), lung squamous cell carcinoma(d), hepatocellular carcinoma (e), and cirrhotic liver nodules (f). Arrows in (d) delineate positive neoplastic cells. FGF19, fibroblastgrowth factor 19; FGFR4, FGF receptor 4; HCC, hepatocellular carcinoma.

HC

T11

6

Col

o205

Col

o201

SW

620

SW

480

HT

29

25 kDa

20

15

100 kDaFGFR4

FGF19

0

10000

20000

30000

40000

50000

60000

70000

80000

FGF19

FGFR4

Rel

ativ

e E

xpre

ssio

n

Colon BreastLiver

Col

o205

SW

620

HT

29

HC

T11

6

SW

480

Col

o320

SN

U18

5

SN

U39

8

SN

U45

3

SN

U47

S

SN

U38

7

MC

F7

Col

o201

Colo205 Colo201

SW620 HCT116

S

T

S

T T

T

a

b

c

Figure 2 FGF19 and FGFR4 expression in cell lines andxenograft tissue. (a) Fgf19 and fgfr4 mRNA expression in a panelof cancer cell lines evaluated by semi quantitative reversetranscription (RT)–PCR. (b) Western blot of FGF19 and FGFR4protein expressed in a panel of colon cancer cell lines. (c) FGF19immunostaining of a subset of colon cancer cell lines xenograftedsubcutaneously into nude mice and excised after 3 weeks(T¼ tumor; S¼ stroma). Images of representative fields at � 400with inset at � 600 are shown (bar¼ 25mm). FGF19, fibroblastgrowth factor 19; FGFR4, FGF receptor 4.

FGF19 and carcinomas in humansLR Desnoyers et al

87

Oncogene

neoplastic epithelial cells but not in the associated mousestroma or adjacent normal tissue. SW620 and HCT116tumors showed FGF19 immunoreactivity in scatteredneoplastic cells. These results confirm that coloncancer cell lines express FGF19 upon growing in vitroin culture dishes as well as in vivo in a subcutaneousxenograft setting.

FGF19 receptor bindingTo examine FGF19’s binding specificity more comple-tely, we assessed its interaction with all known humanFGFRs in their different alternatively spliced formsincluding the recently identified FGFR5 (FGFR1L)(Sleeman et al., 2001). In a solid-phase assay, FGF19dose-dependent binding was restricted to FGFR4(Figure 3a). In a receptor pull-down assay, FGF1 boundto all FGFRs, whereas FGF19 interaction was limitedto FGFR4 (Figure 3b). The dissociation constant ofFGF19 binding to FGFR4 determined using Scatchardanalysis indicated a high-affinity interaction, with Kd

0.25 nM (Figure 3c). These results are consistent withprevious findings (Ornitz et al., 1996; Xie et al., 1999)and further emphasize the unique binding specificityof FGF19 for FGFR4.

Generation of an FGF19-blocking antibodyTo develop an FGF19-neutralizing agent we generated apanel of monoclonal antibodies. From this panel, three

antibodies (1A1, 1A6 and 1D1) were characterizedfurther. In enzyme-linked immunosorbent assays, 1A1and 1A6 bound to FGF19 with a comparable EC50 of40 pM, whereas 1D1 bound with an EC50 of 400 pM(Figure 4a). In a solid-phase receptor binding assay, 1A6blocked FGF19 binding to FGFR4 with an IC50 of 3 nM(Figure 4b). 1A1, 1D1 and an irrelevant controlantibody did not inhibit this interaction.We then analysed antibody neutralization of FGF19-

induced FGF pathway activation in HCC Hep3B cells(Schlessinger, 2004). 1A6 inhibited FGF19-inducedFRS2 and mitogen-activated protein kinase (MAPK)phosphorylation at all doses tested, whereas 1D1 andthe control antibody did not show significant inhibitoryactivity (Figure 4c).FGF19 plays a role in cholesterol homeostatis by

repressing hepatic expression of cholesterol-7-a-hydro-xylase 1 (Cyp7a1), the rate-limiting enzyme for choles-terol and bile acid synthesis (Holt et al., 2003; Yu et al.,2005). To evaluate the FGF19 antibodies further, weexamined their effect on FGF19-induced Cyp7a1 down-regulation in the hepatoblastoma HEPG2 cell line. Inthe absence of antibody, FGF19 reduced Cyp7a1expression by 75% (Figure 4d). The addition of 1.1 mg/ml 1A6 abolished FGF19 activity completely, whereas1A1 reduced this activity only by 50% at 10 mg/ml butnot at lower antibody concentrations. Treatment with acontrol antibody did not affect FGF19 activity.

0

1

2

3

4

0.001 0.01 0.1 1

[FGF19] (µg/ml)

Bin

ding

(O

.D. 4

50nm

)

FGF19

1αIII

b

1βIII

b

1αIII

c

1βIII

c

2αIII

b

2βIII

b

2αIII

c

2βIII

c

3IIIb

3IIIc

4 4

FGFRs

Heparin + –

0.0 0.5 1.0 1.50

4

8

12

[125I-FGF19] (pM)

Bou

nd (

cpm

× 1

0-3)

Bound (µM)

Bou

nd/F

ree

0 5 10 15 20 250.00

0.04

0.08

0.12

FGF1

a FGFR4FGFR1-FGFR3FGFR5

total specific

b

c

Figure 3 FGF19 binding to FGFR5. (a) FGF19 binds specifically to FGFR4 in a solid-phase assay where FGFR1–4 Fc chimericproteins were immobilized and binding was detected using a biotinylated FGF19-specific polyclonal antibody. (b) A receptor pull-down assay demonstrated FGF1 bound to all FGFRs whereas FGF19 interaction was limited to FGFR4. (c) Scatchard analysis of 125I-FGF19 binding to solid phase captured FGFR4-IgG. FGF, fibroblast growth factor; FGFR, FGF receptor.

FGF19 and carcinomas in humansLR Desnoyers et al

88

Oncogene

Because FGFR4 plays a role in cell migration, weevaluated the chemotactic activity of FGF19 (Wanget al., 2004; Spinola et al., 2005). In a modified Boydenchamber assay, FGF19 promoted migration of HCT116cells in a dose-dependent fashion, reaching a maximumat 16 ng/ml (Supplementary Figure S1). 1A6 inhibitedFGF19-induced cell migration (Figure 4e). At higherconcentrations, 1A6 reduced cell migration to 20%below the basal level, likely by inhibiting both exogen-ously added and endogenously produced FGF19.Together, these results demonstrate that 1A6 is a potentinhibitor of FGF19 activity in vitro.

1A6 epitope is localized in the FGF19-binding interfacewith FGFR4To characterize further the FGF19 antibodies welocalized their epitopes and evaluated whether FGF19conformational components contribute to their binding.We first attempted to isolate an epitope-containingpeptide from an FGF19 tryptic digest using an agarose-coupled 1A6 affinity matrix. This approach wasunsuccessful possibly because the FGF19 fragmentationcompromised the conformational integrity of 1A6epitope. We therefore modified the procedure andbound FGF19 to the agarose-coupled antibody before

0

50

100

150

0.1 1 10

[MAbs] (µg/ml)

Bin

ding

(%

Con

trol

)

0

25

50

75

100

Rel

ativ

e C

yp7α

1 E

xpre

ssio

n (%

)

[Abs] (µg/ml) 103.30.120.0400 1.10.3

FGF19

0

1

2

3

4

5

0.0001 0.01 1

[MAbs] (µg/ml)

Bin

ding

(O

.D. 4

50 n

m)

FGF19 - + - + + + + +++

Antibodies Control 1A6 1D1–

p-FRS2

MAPK

p-MAPK

0

50

100

150

200

250

FGF19

Mig

ratio

n (n

umbe

r of

cel

ls)

510.1

1A6 (µg/ml)

1A11A61D1Ctl Ab

1A61A1Ctl Ab

1A11D1

1A6Ctl Ab

p = 0.012p = 0.0055

p = 0.0007

a b

c d

e

Figure 4 1A6 inhibits FGF19 biological activities. (a) Binding of anti-FGF19 monoclonal antibodies to FGF19. (b) Inhibition ofFGF19 binding to solid phase captured FGFR4-Fc by anti-FGF19 monoclonal antibody 1A6. (c) Western blot analysis of FGF19-induced FRS2 and MAPK phosphorylation in HEP3B cells in the presence of anti-FGF19 monoclonal antibodies. (d) Inhibition ofFGF19-induced repression of CYP7a1 expression in HEPG2 cells by monoclonal antibody 1A6. CYP7a1 expression was evaluatedby semi-quantitative reverse transcription (RT) –PCR using gene-specific primers and probes and normalized to GAPDH expression.(e) Inhibition of FGF19-induced cell migration by anti-FGF19 monoclonal antibody 1A6 using a Boyden chamber transwell migrationassay. FGF19, fibroblast growth factor 19; FGFR4, FGF receptor 4; CYP7a1, cholesterol-7-a-hydroxylase 1.

FGF19 and carcinomas in humansLR Desnoyers et al

89

Oncogene

trypsin digestion. Analysis of the total digest demon-strated a complete fragmentation of the adsorbedFGF19 without 1A6 masking of trypsin cleavage sites(Figure 5a). Bound peptides were eluted, identified bymass spectrometry and sequenced. We identified thenon-specifically adsorbed peptides using an irrelevantcontrol antibody coupled to agarose in a parallelexperiment. The agarose-coupled 1A6 specifically iso-lated the FGF19 peptide G133-R155 (Figure 5b andSupplementary Figure S2a). Using agarose-coupled 1A1,we specifically isolated the FGF19 peptide G156-R180(Figure 5c and Supplementary Figure S2b). Becauseagarose conjugation of 1D1 abolished its binding toFGF19, we used a peptide competition assay to identifyits epitope. Only a peptide including the nine mostC-terminal FGF19 amino acids competed 1D1 bindingto the solid-phase adsorbed antigen (Figure 5d). Thispeptide did not compete for 1A1 binding to FGF19

(data not shown). Because overlapping peptides wereused in this competition assay, we surmise that the 1D1epitope is located in the last four FGF19 amino acids(SFEK).We mapped the antibody epitopes on the previously

described structural model of FGF19 bound to FGFR4(Harmer et al., 2004). The epitopes of 1A1 and 1D1 arelocated in a distal portion of FGF19, which is notrepresented on this model due to its lack of orderedstructure. However, the 1A6 epitope is localized in theFGF19-binding interface with FGFR4 (Figure 5e).These results suggest that 1A6 directly occludes thereceptor-binding site of FGF19.

1A6 inhibits the growth of human colon cancer xenograftsTo determine whether FGF19 neutralization couldinhibit tumor growth in vivo, we selected the coloncancer cell lines HCT116 and Colo201, both of which

d

c

a

D173-E182V178-P187

A183-G192

1D1

1A1

1A6

e

Non-specific

0.1 101 10001000.01

Bin

ding

(O

.D.4

50 n

m)

[Peptides] (µg/ml)

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Non-specific

b

[T15

6-18

0]

[T15

6-18

0][T13

3-15

5][T

133-

155]

2665.9

[T13

3-15

5]

[T15

6-18

0]

3240.3

% In

tens

ity%

Inte

nsity

% In

tens

ity

Mass (m/z)

Figure 5 Identification of epitopes recognized by FGF19 antibodies. (a) Mass spectroscopic analysis of soluble fraction of trypsin-digested FGF19 bound to agarose-coupled 1A6. (b) Mass spectroscopic analysis of FGF19 tryptic peptide eluted from agarose-coupled1A6. (c) Mass spectroscopic analysis of FGF19 tryptic peptide eluted from agarose-coupled 1A1. (d) Peptide competition assay toidentify the 1D1 epitope. (e) Mapping of 1A6 epitope (orange) onto the FGF19–FGFR4 interaction model. FGFR4 surface is shownin blue and pink whereas FGF19 ribbon representations are in green and yellow. FGF19, fibroblast growth factor 19; FGFR4, FGFreceptor 4.

FGF19 and carcinomas in humansLR Desnoyers et al

90

Oncogene

express FGF19 and FGFR4 (Figure 2a) and formtumors in vivo. We treated mice with establishedHCT116 xenograft tumors twice weekly with 5mg/kgof either 1A6 or a control antibody. At day 35, 1A6suppressed tumor growth by 57% (P¼ 0.07, n¼ 5)compared to the control group (Figure 6a). We repeatedthis study with a higher dose of 1A6 (15mg/kg; 2� /week) and a statistically significant suppression of tumorgrowth was observed (60% growth inhibition, P¼ 0.01,n¼ 5). To verify that 1A6 inhibited tumor growth byblocking FGF19 activity, we examined tumors at theend of the study for markers of FGFR4 signaling.Activation of FGFR4, FRS2, extracellular signal-regulated kinase (ERK) and b-catenin was significantlydecreased in xenograft tumors from animals treated with1A6 compared to the control antibody (Figure 6b). Inone case, the control-treated xenograft tumor showedlow pFRS2. The relative lack of FRS2 phosphorylationcorrelates with the small tumor size and is consistentwith our hypothesis that this pathway is important fortumor growth.Next, we used xenografts of Colo201, a colon cancer

cell line expressing higher FGF19 levels than HCT116(Figure 2a). 1A6 treatment (30mg/kg; 2� /week)significantly suppressed the growth of establishedtumors (at day 27, 64% growth inhibition, P¼ 0.03,n¼ 5) compared to the control antibody (Figure 6c).Analysis of the excised tumors showed that 1A6significantly decreased FGFR4, FRS2, ERK andb-catenin activation compared to the control antibody(Figure 6d). Together, these results demonstrate 1A6anti-tumor efficacy in colon cancer models and associateits activity with inhibition of FGF19-dependentFGFR4, FRS2, ERK and b-catenin activation.

1A6 prevents hepatocellular carcinomas in FGF19transgenic miceWe demonstrated previously that overexpression ofFGF19 in the skeletal muscle results in developmentof HCCs by 10–12 months of age (Nicholes et al., 2002).To confirm that FGF19 is acting as a tumor promoterin this model, we treated the FGF19 transgenic micewith a tumor initiator, diethylnitrosamine (DEN),which accelerated tumor formation by 50% (unpub-lished data). To determine whether 1A6 could preventHCCs, DEN-accelerated FGF19 transgenic mice weretreated with either 1A6 or control antibody (gp120)for 6 months. At the end of the treatment, all of thecontrol-treated mice had grossly evident multifocal,large HCCs throughout the liver lobes, whereas 1A6-treated animals had either no liver tumors or, in onecase (#1862), a single small tumor present on thediaphragmatic surface of the median lobe (Figure 7a).Liver weights from 1A6-treated mice (mean¼ 1.7170.05g)were significantly lower than liver weights from control-treated mice (mean¼ 3.1570.58g; P¼ 0.014), but werenot significantly different from those of normal FVB wild-type mice (mean¼ 1.5670.08g; P¼ 0.82). In addition,tumor volume was determined by micro-computedtomography (micro-CT) image analysis, corrected fortumor volume with normal FVB liver, and graphed as

a percent of total liver volume (Figure 7b). Percent tumorvolume of 1A6-treated mice (mean¼ 7.573.2%) wassignificantly lower than control gp120-treated mice(mean¼ 23.876.8%; P¼ 0.05). Furthermore, tumorweights strongly correlated with percent tumor volume(r2¼ 0.993702). These data clearly demonstrate that 1A6effectively neutralizes circulating FGF19 to prevent tumorformation in FGF19 transgenic mice.

Discussion

Systemic FGF19 expression in transgenic mice promoteshepatocellular carcinogenesis (Nicholes et al., 2002).However, no report as yet has confirmed FGF19 inhuman cancer. Our results show that FGF19 isexpressed in several human tumors. Neoplastic cells ofcolon adenocarcinoma, lung squamous cell carcin-oma and HCC highly expressed FGF19. FGF19 wasalso detected in cirrhosis, a preneoplastic conditionthat often leads to liver carcinoma (Anthony, 1979;Kobayashi et al., 2006). These data raise the possibilitythat FGF19 may be involved in tumor pathogenesisand, particularly for the liver, in early neoplasticprogression.Previous work suggested that FGF19 binds selectively

to FGFR4 (Xie et al., 1999). FGFR4 deregulation hasbeen linked to cancer pathogenesis. FGFR4 truncationand cytoplasmic localization correlates with malignancyin pituitary adenomas (Qian et al., 2004). FGFR4expression predicts short survival in astrocytoma(Yamada et al., 2002). A single nucleotide polymorph-ism at codon 388 of FGFR4 strongly correlates with theprognosis of prostate and lung adenocarcinoma, headand neck squamous cell carcinoma and soft tissuesarcoma (Morimoto et al., 2003; Streit et al., 2004;Wang et al., 2004). Our data demonstrate that neo-plastic cells of colon adenocarcinoma, lung squamouscell carcinoma and HCC often express FGFR4 inconjunction with FGF19, suggesting that their interac-tion may contribute to tumor progression.The importance of FGF19 in tumor growth was

further examined by generating anti-FGF19 monoclonalantibodies. Among these, 1A6 blocked the interaction ofFGF19 with FGFR4. In HCC HEP3B cells, 1A6inhibited FGF19 stimulation of FGFR4-mediatedevents including FGFR4, FRS2 and MAPK phosphor-ylation and CYP7a1 gene repression. 1A6 also effec-tively inhibited FGF19 chemotactic activity towardHCT116 cells.To define how 1A6 inhibits FGF19 at the molecular

level, we identified and compared the epitopes of threeFGF19 antibodies. 1A1 recognized an epitope locatedbetween the amino acids 156–180, whereas 1D1 boundto the most C-terminal five residues. The regionsrepresented by these epitopes are absent from therecently reported crystal structure of FGF19 (Harmeret al., 2004). The 1A6-binding epitope mapped to theb12 strand of FGF19. Based on our structural model ofthe FGF19–FGFR4 complex, the b12 strand is locatedon the surface of the ligand’s interface with the receptor

FGF19 and carcinomas in humansLR Desnoyers et al

91

Oncogene

HCT116 Xenografta

b

Colo201 Xenograft

0 5 10 15 20 25 300

200

400

600

800

Treatment Days

Tum

or V

olum

e (m

m3 )

Control Ab

1A6

15 20 25 30 350

200

400

600

800

1000

1200

Treatment Days

Tum

or V

olum

e (m

m3 )

Control Ab

1A6

d

1398

737

785

181

223

89

Tumor Volume

pERK

non-phosph. β-catenin

β-catenin

pFGFR4

pFRS2

gp120 1A6

HCT116 Xenograft

ERK

pERK

β-catenin

pFGFR4

FGFR4

Tumor Volume

ERK

1308

253

398

389

197

925

gp120 1A6

Colo201 Xenograft

pFRS2

non-phosph. β-catenin

FRS2

FGFR4

FRS2

c

Figure 6 Effect of 1A6 on colon tumor xenografts. Athymic mice were subcutaneously inoculated with 5� 106 HCT116 (a) orColo201 (c) cells. Mice bearing established tumors of equivalent volumes (B100mm3) were randomized into groups and treatedintraperitoneally twice weekly with 1A6 or a control antibody. Results are given as mean tumor volume 7s.e.m. At the end of thestudies HCT116 xenograft tumors (b) and Colo201 xenograft tumors (d) from 1A6 or control antibody-treated mice were excised,homogenized and analysed for FGFR4, FRS2, ERK and b-catenin activation by western blot. ERK, extracellular signal-regulatedkinase; FGFR4, FGF receptor 4; FRS2, fibroblast growth factor receptor substrate 2.

FGF19 and carcinomas in humansLR Desnoyers et al

92

Oncogene

(Figure 6e). Thus, 1A6 directly competes with FGFR4for FGF19 interaction.We then tested the anti-tumor efficacy of 1A6 in vivo

using colon cancer xenograft and FGF19 transgenicHCC models. 1A6 treatment significantly reducedgrowth of HCT116 and Colo201 xenografts comparedto a control antibody. In conjunction, 1A6 substantiallyreduced FGFR4, FRS2, ERK and b-catenin activation,suggesting that FGF19 plays a role in the growth ofthese tumors. Since FGF19 does not have a directmurine ortholog that is bound by 1A6, the anti-tumoractivity of 1A6 is likely mediated by neutralization oftumor-produced FGF19. In addition, the ability of 1A6to prevent HCCs in FGF19 transgenic mice whereFGF19 is ectopically expressed in the skeletal musclefurther confirms that 1A6 effectively neutralizes theoncogenic activity of FGF19.The FGF19 transgenic hepatocellular tumors have a

high frequency of b-catenin nuclear localization andactivating mutations (Nicholes et al., 2002). b-Cateninpathway deregulation promotes tumor development in

several types of cancer, including colorectal adenocarci-noma and HCC. Activation of this pathway impactsneoplastic cell proliferation, epithelial mesenchymaltransition, invasion and metastasis. Repression ofb-catenin activation by 1A6 in colon tumor xenograftsprovides further evidence for a link between FGF19 andwnt pathway activity in the context of cancer.FGF19 is potentially an attractive target for cancer

therapy for several reasons. FGF19 expression in theadult is restricted to the gall bladder (Xie et al., 1999)and is specific in cancer tissues to the neoplastic cells.FGF19 binds selectively to FGFR4, which has beenassociated with cancer progression. Knockout micedeficient in FGFR4 or the FGF19 murine ortholog,FGF15, develop normally and do not show grossabnormalities, suggesting that inhibition of FGF19may not have significant safety limitations . Finally,the activity of the FGF19-FGFR4 axis in colorectal andperhaps other tumors may provide useful markers foridentifying patients who could benefit most from anti-FGF19 therapy.

Control Ab treated 1A6 treated

b Control Ab treated

1A6 treated

Total Tumor Volume to Total Liver Volume

70

60

50

40

30

20

10

0

-101A6 control Ab

Treatment

Per

cent

age

Tum

or V

olum

e

to T

otal

Liv

er V

olum

e (%

)

a

Figure 7 Effect of 1A6 on FGF19 transgenic hepatocellular carcinomas. (a) FGF19 transgenic mice were treated with either a controlantibody or 1A6 and liver was collected for gross evaluation. (b) Micro-CT analysis using an iodinated triglyceride for enhancement ofnormal hepatocytes demonstrates increased unenhanced tumor volume in control-treated versus 1A6-treated liver. FGF19, fibroblastgrowth factor 19; micro-CT, micro-computed tomography.

FGF19 and carcinomas in humansLR Desnoyers et al

93

Oncogene

Materials and methods

Gene expressionTotal RNA from frozen tissue samples was extracted usingRNA STAT-60 according to the manufacturer’s protocol(Tel-test ‘B’ Inc., Friendswood, TX, USA). Total RNA fromcultured cells was isolated with RNeasy kit using themanufacturer’s protocol (Qiagen, Valencia, CA, USA), treatedwith DNAse-I (Ambion, Austin, TX, USA) and used for real-time PCR. Specific primers and fluorogenic probes for humanFGF19, FGFR4 and RPL19 (Supplemental Data Table 1)were designed using Primer Express 1.1 (PE, AppliedBiosystems, Foster City, CA, USA). Gene-specific signals,evaluated in triplicate, were normalized to the RPL19.

In situ hybridization33P-UTP-labeled sense or antisense probes corresponding tohuman FGFR4 (nucleotides 435–1183 of NM_022963) orFGF19 (nucleotides 495–1132 of NM_005117) were gener-ated by PCR (Mauad et al., 1994). Sections were processedfor in situ hybridization as described previously (Jubb et al.,2006).Tissue microarrays were obtained from the following

sources: multi-tumor TMA was made at Genentech Inc.,HCC TMAs were from Pentagen (Seoul, Korea) (catalog nos.A217 and A204) or Cybrdi (Frederick, MD, USA) (catalognos. CC03-01-003 and CC03-02-01-001).

ImmunohistochemistryFormalin-fixed paraffin-embedded tissue sections were treatedfor antigen retrieval using Trilogy (Cell Marque, Rocklin, CA,USA) and then incubated with 10 ug/ml FGF19 antibody(1D1; Genentech Inc.). The immunostaining was accomplishedusing a biotinylated secondary antibody, an ABC-HRPreagent (Vector Labs, Burlingame, CA, USA) and a metal-enhanced DAB colorimetric peroxydase substrate (PierceLaboratories, Rockford, IL, USA).

Immunoprecipitation and immunoblottingTotal proteins from cultured cells were extracted on ice for30min in extraction buffer (20mM Tris, pH 8, 137mM NaCl,1mM ethylene glycol tetraocetic acid (EGTA), 1% Triton X-100, 10% glycerol, 1.5mM MgCl2, complete proteaseinhibitor cocktail (Roche Applied Sciences, Indianapolis,IN, USA)). Lysates (100 mg protein) were incubated in 1mlPBS/0.1% Triton with agarose-coupled antibodies for 1 h at41C, then washed with buffer and eluted with 10 ml Elutionbuffer (Pierce Biotechnology, Rockford, IL, USA) andanalysed by immunoblot using biotinylated FGF19 antibody(BAF969; R&D Systems, Minneapolis, MN, USA), FGFR4antibody (1G7; Genentech Inc.) and IRDye 800-conjugatedsecondary reagents and visualized using the Odyssey scanner(Li-Cor Biotechnology, Lincoln, NE, USA).

Solid-phase receptor binding assayMaxisorb 96-well plates were coated overnight at 41C with50 ml of 2mg/ml anti-human immunoglobulin Fcg fragmentspecific (Jackson Immunoresearch, West Grove, PA, USA)and used to capture 1mg/ml FGFR-Fc chimeric proteins(R&D Systems). After blocking with phosphate-bufferedsaline (PBS)/3% bovine serum albumin (BSA), FGF19 andheparin were added and incubated 2 h (Seikagaku Corpora-tion, Tokyo, Japan). A biotinylated FGF19-specific polyclonalantibody (BAF969; R&D Systems) followed by streptavidin-HRP and TMB was used for detection.

Receptor pull-down assayFGFR-Fc chimeric proteins (400 ng) were incubated withFGF19 or FGF1 and heparin (200 ng) in 50:50 Dulbecco’smodified essential media Ham F12 containing 10mM HEPES,pH 7.4, and 0.1% BSA for 1 h followed by Protein G-agarose(20ml) for 30min. The matrix was washed with PBS/0.1%Triton X-100, eluted with sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE) sample buffercontaining reducing agent and analysed by western blottingusing biotinylated FGF19 antibody (BAF969) or biotinylatedFGF1 antibody (BAF232; R&D Systems).

Generation of FGF19 monoclonal antibodiesBalb/c mice were immunized with FGF19-His. Spleens wereharvested after 8 weeks and hybridomas were generated.Selected FGF19 antibody producing hybridomas were sub-cloned twice to insure monoclonality. The clones were ino-culated for ascites production and antibodies were purified byProtein A-agarose affinity chromatography.

Solid-phase antibody binding assayNon-specific binding sites of HisGrab nickel-coated plates(Pierce) were saturated with PBS/3% BSA, and then incubatedwith 1 mg/ml FGF19-His for 1 h. The plates were washed andincubated for 1 h with FGF19 antibodies in the presence or theabsence of FGF19 peptides and detected using a HRP-conjugated anti-mouse IgG (Jackson Immunoresearch) andthe TMB peroxidase colorigenic substrate (KPL, Gaithers-burg, MD, USA).

CYP7a1 expression analysisHEPG2 cells were starved overnight in serum-free media andtreated with 100 ng/ml FGF19 for 6 h in the presence orabsence of antibodies. CYP7a1 expression was evaluated intriplicate by semi-quantitative reverse transcription (RT)–PCR(Taqman ABI PRISM 7700, Applied Biosystems) and normal-ized to GAPDH expression.

FGFR4/MAPK phosphorylationHEP3B cells starved overnight in serum-free media weretreated with 40 ng/ml FGF19 for 10min in the presence or theabsence of antibodies. Cells were lysed in R27A buffer(Upstate, Charlottesville, VA, USA) with 10mM NaF, 1mM

sodium orthovanadate, and Complete protease inhibitor tablet(Roche). Lysates were prepared, electrophoresed and analysedby Immunoblot using phospho-FRS2 (Santa Cruz, SantaCruz, CA, USA), phospho-MAPK and MAPK-specific anti-bodies (Cell Signaling, Danvers, CA, USA).

Cell migration assayCell motility was measured using a modified Boyden chamberassay (Eriksson et al., 1992). Briefly the surface of 8mm porosity24-well format PET membrane filters (BD Biosciences, Bedford,MA, USA) was coated overnight at 41C with 50ml of type 1collagen (50mg/ml; Sigma, St Louis, MO, USA) in 0.02M aceticacid. Cells (5� 104) in serum-free minimal essential mediumcontaining 0.1% BSA were added to the upper chamber. Thelower chamber was filled with the same media and the plates wereincubated at 371C. The next day the upper chamber was wipedwith a cotton swab and the cells that migrated to the lower side ofthe insert were stained and counted under a microscope.Triplicate sets of data were averaged for each condition.

Antibody epitope excisionFGF19 (10mg) was incubated for 2 h in 50mM Tris, pH 7.4with 50 ml of agarose-coupled antibody. The resin was washed

FGF19 and carcinomas in humansLR Desnoyers et al

94

Oncogene

and digested with 0.1mg trypsin (Promega, Madison, WI,USA) overnight at 371C in 100mM ammonium bicarbonate(pH 8). The gel slurry was washed and the bound peptides wereeluted with 10% trifluoroacetic acid and analysed by MALDI-TOF-MS (Voyager; Applied Biosystems). Candidate peptideswere subjected to collision-induced dissociation (QSTAR) andmanually sequenced to confirm the peptide mass mappingidentification (Supplementary Figure S2).

Xenograft experimentsAll animal protocols were approved by an InstitutionalAnimal Care and Use Committee. Six- to eight-week-oldathymic BALB/c female mice (Charles Rivers Inc., Wilming-ton, MA, USA) were inoculated subcutaneously with 5� 106

cells (200ml/mouse). After 7 days, mice bearing tumors ofequivalent volumes (B100mm3) were randomized into groups(n¼ 10) and treated intraperitoneally twice weekly. Tumorswere measured with an electronic caliper (Fowler Sylvac Ultra-Cal Mark III) and average tumor volume was calculated usingthe formula (W2�L)/2, where W is the smaller diameter andL is the larger diameter.

FGFR4, FRS2, ERK and b-catenin phosphorylation inxenograft tumorsTumors excised from treated animals were homogenized inlysis buffer (50mM Tris-HCl, pH 7.5, 150mM NaCl, 1% NP-40, 1mM EDTA, 0.25% sodium deoxycholate, 1mM NaF,1mM sodium orthovanadate and complete protease inhibitor(Roche)). Equal amounts of proteins were incubated with 1mg-specific FGFR4 (1G7; Genentech Inc.) or FRS2 (UpState)antibody immobilized onto Protein A-Sepharose for 2 h at 41Cthen washed with lysis buffer and eluted with 2� Laemmlibuffer, boiled, and microcentrifuged. Immunoblots were donewith phosphotyrosine antibody (4G10, UpState), phospho-ERK2 antibody (Santa Cruz Biotech) or N-terminally dephos-phorylated b-catenin antibody (UpState). Membranes were s-tripped (Pierce) and reprobed with appropriate antibodies todetermine total proteins.

Micro-CT imaging and analysis of hepatocellular carcinomas inFGF19 TG miceLiver tumors were identified by micro-CT imaging withFenestra-LC (Lee et al., 1997). At 6 months of age, FGF19transgenic mice were injected with Fenestra-LC (AdvancedResearch Technologies Inc., Saint-Laurent, Quebec, Canada),20 ml/g intravenously, and a conscious 3-h hepatic uptake wasallowed before mice were euthanized and livers analysed (mCT40 system; Scanco Medical, Bassersdorf, Switzerland). Wholelivers were submerged in soybean oil (Sigma-Aldrich, St Louis,MO, USA) in preparation for micro-CT imaging. Ninety-minscans were obtained at 30-mm isotropic voxel size, with 512projections at an integration time of 300ms, energy of 45 keVand tube current of 177 mA. Volumetric image files wereanalysed using image analysis software from AnalyzeDirect(Lenexa, KS, USA). An intensity threshold of �16 Hounds-field Units (HU) was used to segment the tissue mass from thebackground signal (soybean oil). A second threshold (26 HU)was used for functional hepatic tissue volumes that accumu-lated contrast. An average of the total low-intensity hepaticvolumes from wild-type FVB mice was subtracted to obtainvolumes associated only with tumors. Data are expressed as apercentage of tumor volume of total liver volume.

Statistical analysisStatistical significance was analysed using the unpaired two-tailed Student’s t-test. Values of Po0.05 were consideredsignificant. Data are expressed as the mean7s.e.m.

Acknowledgements

We thank Chris Wiesmann for structural modeling; Robin ETaylor and Julio Ramirez for paraffin embedding andsectioning; Sheila Bheddah for immunohistochemistry assis-tance; and Hartmut Koeppen, Paul Polakis, and Vishva Dixitfor helpful discussions.

References

Anthony PP. (1979). Precancerous changes in the human liver.J Toxicol Environ Health 5: 301–313.

Bange J, Prechtl D, Cheburkin Y, Specht K, Harbeck N,Schmitt M et al. (2002). Cancer progression and tumor cellmotility are associated with the fgfr4 arg(388) allele. CancerRes 62: 840–847.

Cappellen D, De Oliveira C, Ricol D, de Medina S, Bourdin J,Sastre-Garau X et al. (1999). Frequent activating mutationsof fgfr3 in human bladder and cervix carcinomas. Nat Genet23: 18–20.

Chesi M, Brents LA, Ely SA, Bais C, Robbiani DF, Mesri EAet al. (2001). Activated fibroblast growth factor receptor 3 isan oncogene that contributes to tumor progression inmultiple myeloma. Blood 97: 729–736.

Chesi M, Nardini E, Brents LA, Schrock E, Ried T, KuehlWM et al. (1997). Frequent translocation t(4;14)(p16.3;q32.3) in multiple myeloma is associated withincreased expression and activating mutations of fibroblastgrowth factor receptor 3. Nat Genet 16: 260–264.

Dorkin TJ, Robinson MC, Marsh C, Bjartell A, Neal DE,Leung HY. (1999). Fgf8 over-expression in prostatecancer is associated with decreased patient survival andpersists in androgen independent disease. Oncogene 18:2755–2761.

Eriksson A, Siegbahn A, Westermark B, Heldin CH, Claesson-Welsh L. (1992). Pdgf alpha- and beta-receptors activateunique and common signal transduction pathways. EMBO J11: 543–550.

Eswarakumar VP, Lax I, Schlessinger J. (2005). Cellularsignaling by fibroblast growth factor receptors. CytokineGrowth Factor Rev 16: 139–149.

Gowardhan B, Douglas DA, Mathers ME, McKie AB,McCracken SR, Robson CN et al. (2005). Evaluationof the fibroblast growth factor system as a potential targetfor therapy in human prostate cancer. Br J Cancer 92:320–327.

Harmer NJ, Pellegrini L, Chirgadze D, Fernandez-Recio J,Blundell TL. (2004). The crystal structure of fibroblastgrowth factor (fgf) 19 reveals novel features of the fgf familyand offers a structural basis for its unusual receptor affinity.Biochemistry 43: 629–640.

Holt JA, Luo G, Billin AN, Bisi J, McNeill YY, Kozarsky KFet al. (2003). Definition of a novel growth factor-dependentsignal cascade for the suppression of bile acid biosynthesis.Genes Dev 17: 1581–1591.

Holzmann K, Kohlhammer H, Schwaenen C, Wessendorf S,Kestler HA, Schwoerer A et al. (2004). Genomic DNA-chiphybridization reveals a higher incidence of genomic

FGF19 and carcinomas in humansLR Desnoyers et al

95

Oncogene

amplifications in pancreatic cancer than conventional com-parative genomic hybridization and leads to the identificationof novel candidate genes. Cancer Res 64: 4428–4433.

Huang YQ, Li JJ, Nicolaides A, Zhang WG, Friedman-KienAE. (1993). Increased expression of fibroblast growthfactors (fgfs) and their receptor by protamine and suraminon kaposi’s sarcoma-derived cells. Anticancer Res 13:887–890.

Jaakkola S, Salmikangas P, Nylund S, Partanen J, ArmstrongE, Pyrhonen S et al. (1993). Amplification of fgfr4 gene inhuman breast and gynecological cancers. Int J Cancer 54:378–382.

Jang JH, Shin KH, Park JG. (2001). Mutations in fibroblastgrowth factor receptor 2 and fibroblast growth factorreceptor 3 genes associated with human gastric and color-ectal cancers. Cancer Res 61: 3541–3543.

Jang JH, Shin KH, Park YJ, Lee RJ, McKeehan WL,Park JG. (2000). Novel transcripts of fibroblast growthfactor receptor 3 reveal aberrant splicing and activation ofcryptic splice sequences in colorectal cancer. Cancer Res 60:4049–4052.

Jeffers M, LaRochelle WJ, Lichenstein HS. (2002). Fibroblastgrowth factors in cancer: therapeutic possibilities. ExpertOpin Ther Targets 6: 469–482.

Jubb AM, Pham TQ, Frantz GD, Peale Jr FV, Hillan KJ.(2006). Quantitative in situ hybridization of tissue micro-arrays. Methods Mol Biol 326: 255–264.

Kiuru-Kuhlefelt S, Sarlomo-Rikala M, Larramendy ML,Soderlund M, Hedman K, Miettinen M et al. (2000). Fgf4and int2 oncogenes are amplified and expressed in kaposi’ssarcoma. Mod Pathol 13: 433–437.

Kobayashi M, Ikeda K, Hosaka T, Sezaki H, Someya T,Akuta N et al. (2006). Dysplastic nodules frequently developinto hepatocellular carcinoma in patients with chronic viralhepatitis and cirrhosis. Cancer 106: 636–647.

Krejci P, Dvorakova D, Krahulcova E, Pachernik J, Mayer J,Hampl A et al. (2001). Fgf-2 abnormalities in b cell chroniclymphocytic and chronic myeloid leukemias. Leukemia 15:228–237.

Lee Jr FT, Chosy SG, Naidu SG, Goldfarb S, Weichert JP,Bakan DA et al. (1997). Ct depiction of experimentalliver tumors: Contrast enhancement with hepatocyte-selec-tive iodinated triglyceride versus conventional techniques.Radiology 203: 465–470.

Liscia DS, Merlo GR, Garrett C, French D, Mariani-Costantini R, Callahan R. (1989). Expression of int-2 mrnain human tumors amplified at the int-2 locus. Oncogene 4:1219–1224.

Marsh SK, Bansal GS, Zammit C, Barnard R, Coope R,Roberts-Clarke D et al. (1999). Increased expression offibroblast growth factor 8 in human breast cancer. Oncogene18: 1053–1060.

Mattila MM, Ruohola JK, Valve EM, Tasanen MJ, SeppanenJA, Harkonen PL. (2001). Fgf-8b increases angiogeniccapacity and tumor growth of androgen-regulated s115breast cancer cells. Oncogene 20: 2791–2804.

Mauad TH, van Nieuwkerk CM, Dingemans KP, Smit JJ,Schinkel AH, Notenboom RG et al. (1994). Mice withhomozygous disruption of the mdr2 p-glycoprotein gene. Anovel animal model for studies of nonsuppurative inflam-matory cholangitis and hepatocarcinogenesis. Am J Pathol145: 1237–1245.

Morimoto Y, Ozaki T, Ouchida M, Umehara N, Ohata N,Yoshida A et al. (2003). Single nucleotide polymorphism infibroblast growth factor receptor 4 at codon 388 is associatedwith prognosis in high-grade soft tissue sarcoma. Cancer 98:2245–2250.

Nicholes K, Guillet S, Tomlinson E, Hillan K, Wright B,Frantz GD et al. (2002). A mouse model of hepatocellularcarcinoma : Ectopic expression of fibroblast growth factor19 in skeletal muscle of transgenic mice. Am J Pathol 160:2295–2307.

Ornitz DM, Itoh N. (2001). Fibroblast growth factors. GenomeBiol 2: REVIEWS3005.

Ornitz DM, Xu J, Colvin JS, McEwen DG, MacArthur CA,Coulier F et al. (1996). Receptor specificity of the fibroblastgrowth factor family. J Biol Chem 271: 15292–15297.

Popovici C, Zhang B, Gregoire MJ, Jonveaux P, Lafage-Pochitaloff M, Birnbaum D et al. (1999). The t(6;8)(q27;p11)translocation in a stem cell myeloproliferative disorder fusesa novel gene, fop, to fibroblast growth factor receptor 1.Blood 93: 1381–1389.

Powers CJ, McLeskey SW, Wellstein A. (2000). Fibroblastgrowth factors, their receptors and signaling. Endocr RelatCancer 7: 165–197.

Qian ZR, Sano T, Asa SL, Yamada S, Horiguchi H, Tashiro Tet al. (2004). Cytoplasmic expression of fibroblast growthfactor receptor-4 in human pituitary adenomas: Relation totumor type, size, proliferation, and invasiveness. J ClinEndocrinol Metab 89: 1904–1911.

Richelda R, Ronchetti D, Baldini L, Cro L, Viggiano L,Marzella R et al. (1997). A novel chromosomal trans-location t(4; 14)(p16.3; q32) in multiple myeloma involvesthe fibroblast growth-factor receptor 3 gene. Blood90: 4062–4070.

Ruohola JK, Viitanen TP, Valve EM, Seppanen JA,Loponen NT, Keskitalo JJ et al. (2001). Enhanced invasionand tumor growth of fibroblast growth factor 8b-over-expressing mcf-7 human breast cancer cells. Cancer Res61: 4229–4237.

Schlessinger J. (2004). Common and distinct elements incellular signaling via egf and fgf receptors. Science 306:1506–1507.

Shimokawa T, Furukawa Y, Sakai M, Li M, Miwa N,Lin YM et al. (2003). Involvement of the fgf18 gene incolorectal carcinogenesis, as a novel downstream targetof the beta-catenin/t-cell factor complex. Cancer Res 63:6116–6120.

Sleeman M, Fraser J, McDonald M, Yuan S, White D,Grandison P et al. (2001). Identification of a new fibroblastgrowth factor receptor, fgfr5. Gene 271: 171–182.

Spinola M, Leoni V, Pignatiello C, Conti B, Ravagnani F,Pastorino U et al. (2005). Functional fgfr4 gly388argpolymorphism predicts prognosis in lung adenocarcinomapatients. J Clin Oncol 23: 7307–7311.

Streit S, Bange J, Fichtner A, Ihrler S, Issing W, Ullrich A.(2004). Involvement of the fgfr4 arg388 allele in headand neck squamous cell carcinoma. Int J Cancer 111:213–217.

Wang J, Stockton DW, Ittmann M. (2004). The fibroblastgrowth factor receptor-4 arg388 allele is associated withprostate cancer initiation and progression. Clin Cancer Res10: 6169–6178.

Xiao S, Nalabolu SR, Aster JC, Ma J, Abruzzo L, Jaffe ESet al. (1998). Fgfr1 is fused with a novel zinc-finger gene,znf198, in the t(8;13) leukaemia/lymphoma syndrome. NatGenet 18: 84–87.

Xie MH, Holcomb I, Deuel B, Dowd P, Huang A, Vagts Aet al. (1999). Fgf-19, a novel fibroblast growth factor withunique specificity for fgfr4. Cytokine 11: 729–735.

Yamada SM, Yamada S, Hayashi Y, Takahashi H, TeramotoA, Matsumoto K. (2002). Fibroblast growth factor receptor(fgfr) 4 correlated with the malignancy of human astro-cytomas. Neurol Res 24: 244–248.

FGF19 and carcinomas in humansLR Desnoyers et al

96

Oncogene

Yu C, Wang F, Jin C, Huang X, McKeehan WL. (2005).Independent repression of bile acid synthesis and activationof c-jun n-terminal kinase (jnk) by activated hepatocytefibroblast growth factor receptor 4 (fgfr4) and bile acids.J Biol Chem 280: 17707–17714.

Zaharieva BM, Simon R, Diener PA, Ackermann D,Maurer R, Alund G; et al. (2003). High-throughputtissue microarray analysis of 11q13 gene amplification(ccnd1, fgf3, fgf4, ems1) in urinary bladder cancer. J Pathol201: 603–608.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

FGF19 and carcinomas in humansLR Desnoyers et al

97

Oncogene