Ligand-independent activation of fibroblast growth factor receptor-2 by ...

3 downloads 0 Views 1MB Size Report
tyrosine kinases, the fibroblast growth factor receptors. (FGFRs). ... mutagenesis mediated FGFR2 activation in assays for transforming activity and anchorage- ...
Oncogene (1997) 15, 817 ± 826  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Ligand-independent activation of ®broblast growth factor receptor-2 by carboxyl terminal alterations Matthew V Lorenzi, Paola Castagnino, Qiong Chen, Marcio Chedid and Toru Miki Laboratory of Cellular and Molecular Biology, National Cancer Institute, Bethesda, Maryland 20892, USA

To assess the e€ect(s) of the C-terminal domain on FGFR2 function, we engineered a series of mutant FGFR2 cDNAs encoding deletions in the C-terminus of the receptor and compared their growth properties in NIH3T3 ®broblasts. In contrast to FGFR2-WT, receptors with C-terminal truncations induced ligandindependent transformation of NIH3T3 cells and transfectants expressing these mutant receptors eciently formed colonies in semisolid medium. Introduction of point mutations (Y to F) into the C-terminus of FGFR2 at positions 813, 784 or 780 revealed that these mutant receptors also displayed activities similar to that of C-terminally truncated receptors. C-terminally altered FGF receptors did not show an increase in the basal level of receptor phosphorylation compared to that of FGFR2WT suggesting that elevated receptor phosphorylation does not underlie the transforming activity of these receptors. Interestingly, expression of transforming FGFR2 derivatives, unlike H-Ras transformed cells, did not result in the activation of the mitogen-activated protein kinases (MAPKs), p42/ERK2 and p44/ERK1, indicating that this pathway is not constitutively active in FGFR2-transformed cells. Finally, we report the overexpression of FGFR2 mRNA and protein in several human tumor cell lines suggesting activation of the receptor in these tumors. Keywords: FGFs; cell transformation; phosphotyrosine; receptor tyrosine kinases; MAP kinases

Introduction Members of the ®broblast growth factor (FGF) family are involved in a variety of physiological and pathological processes including mitogenesis, organogenesis, wound healing and angiogenesis (for reviews see Wilkie et al., 1995; Basilico and Moscatelli, 1992). To date, ten structurally related FGFs, aFGF/FGF-1, bFGF/FGF-2, INT-2/FGF-3, HST(k-FGF)/FGF-4, FGF-5, HST-2/FGF-6, KGF/FGF-7, AIGF/FGF-8, GAF/FGF-9 and FGF-10 have been identi®ed. The FGFs elicit their biological responses through high anity binding and activation of a family of receptor tyrosine kinases, the ®broblast growth factor receptors (FGFRs). Four members of structurally related FGFRs (FGFR1/Flg, FGFR2/Bek, FGFR3 and FGFR4) have been characterized (Dionne et al., 1990; Miki et al., 1991; Keegan et al., 1991; Partanen et al., 1991; Ron et al., 1993). Extracelluarly, the Correspondence: MV Lorenzi Received 28 February 1997; revised 30 April 1997; accepted 1 May 1997

receptors are composed of two or three immunoglobulin (Ig)-like domains which mediate ligand binding and interaction with heparin proteoglycans (Kan et al, 1993; Gao and Goldfarb, 1995). Intracellularly, the receptors contain a juxtamembrane domain, a split tyrosine kinase domain separated by kinase insert sequences and a carboxyl terminal domain containing multiple tyrosine phosphorylation sites. The diversity of the FGFR family is increased further by alternative RNA splicing which generates structural isoforms of FGFRs that vary in their extracellular and/or intracellular domains (ChampionArnaud et al, 1991; Miki et al., 1992; Chellaiah et al., 1994). Structural variants of FGFRs di€ering in extracellular domains exhibit distinct ligand-binding speci®city and anity (Miki et al., 1992; Chellaiah et al., 1994). The ®rst Ig domain of FGFR1 and FGFR2 has been demonstrated to be dispensable for binding of aFGF and bFGF (Mansukhani et al., 1990; Miki et al., 1992). However, the importance of this domain was suggested by a report describing the switch in expression of a FGFR2 isoform with three Ig domains in normal brain to a two Ig domain isoform in malignant astrocytomas (Yamaguchi et al., 1994). While the second Ig domain of FGFR2 has been shown to be important for aFGF binding, the third Ig domain of the receptor governs binding of bFGF or KGF (Miki et al., 1992; Yayon et al., 1992). Interestingly, the KGF receptor (KGFR) and FGFR2 proteins arise from alternative splicing of exons encoding the third Ig domain; the carboxyl terminal half of the third Ig domain of FGFR2 contains sequences responsible for high anity binding of bFGF while the corresponding region in KGFR mediates KGF binding. Moreover, the observation that KGFR expression is tightly restricted to epithelial cells (Miki et al., 1992; Finch et al., 1995), suggests that this FGF-binding speci®city generated by alternative splicing occurs in a cell type-speci®c manner. Alternative splicing also generates isoforms of KGFR/FGFR2 with altered carboxyl termini (Champion-Arnaud et al., 1991; Yan et al., 1993; Itoh et al., 1994). A consistent structural feature of these isoforms is the deletion of a portion of the carboxyl terminal domain containing potential tyrosine phosphorylation sites. In this study we examined the role of the carboxyl terminal domain of FGFR2 on the activity of the receptor in NIH3T3 ®broblasts. We found that deletions of the C-terminal domain of the receptor introduced by alternative splicing or site-directed mutagenesis mediated FGFR2 activation in assays for transforming activity and anchorage-independent growth. These results provide evidence for the importance of the C-terminal domain in regulating receptor activity and suggest that naturally occurring

FGFR2 activation by C-terminal alterations MV Lorenzi et al

818

isoforms of FGFR with C-terminal alterations have enhanced growth-promoting activity.

Results Removal of the C-terminal domain of FGFR2 increases the receptor's transforming activity We have previously described a constitutively active FGFR2 isoform with a variant C-terminal domain from a rat osteosarcoma cell line, ROS 17/2.8 (Lorenzi et al., 1996). This receptor, designated FGFR2-ROS, was generated by a chromosomal rearrangement with a novel gene, FRAG1, which replaces the normal Cterminal domain of the receptor. When the FRAG1 sequence from FGFR2-ROS was removed, the resulting receptor, FGFR2D7647, still showed partial activity in the absence of ligand suggesting that the Cterminal domain can negatively regulate receptor

activity. To examine this possibility in more detail, we constructed a series of deletion derivatives of FGFR2 which were used to transfect NIH3T3 cells. Transfection of vector alone or DNA encoding wildtype FGFR2 (FGFR2-WT) did not induce signi®cant transforming activity in NIH3T3 cells (Figure 1a). In contrast, KGFR, a splicing variant of FGFR2 containing a di€erent third Ig domain and can be activated by KGF secreted from NIH3T3 cells, induced high tittered transforming activity. This activity was equivalent to that of PDGF-BB, which results in constitutive activation of the PDGFR-b pathway in NIH3T3 cells (data not shown). On the other hand, receptors containing deletions in their C-terminal domains showed modest transforming activity. Three of these deletions, D7647 and D7767 and D7897 induced foci relatively well with transforming activities similar to that observed for FGFR2-ET (0.66103 f.f.u./pmol DNA), an alternatively spliced FGFR2 containing a partially truncated C-terminal

a

b pCEV29

KGFR

FGFR2-WT

FGFR2-ET

∆802-

∆789-

∆776-

∆764-

7Figure 1 Transforming activity and soft agar growth of FGFR2 mutants containing carboxyl terminal truncations. (a) Transforming activity of FGFR2 derivatives in NIH3T3 ®broblasts. Mutant FGFR2 cDNAs were constructed to encode receptors with nested deletions in their C-terminal domain. Dashed lines indicate the removal of these residues from the wild type C-terminus while open boxes mark residues remaining in the domain. Numbers above the sequence denote the location of C-terminal tyrosine residues. FGFR2-ET is an alternatively spliced isoform encoding a partially truncated C-terminus and four amino acid substitutions (shown within the open box). NIH3T3 cells were transfected by calcium phosphate precipitation using serial dilutions (0.01 ± 1.0 mg) of the indicated plasmid DNAs and 40 mg of calf thymus DNA as carrier. Transforming activity was expressed as the number of foci produced by each DNA normalized to the pmol of the DNA expression construct used (focus forming unit (f.f.u.)/pmol DNA). KGFR and pCEV29 were used as positive and negative controls, respectively, in these assays. The data presented represent the average of four independent experiments. (b) Growth of NIH3T3 transfectants expressing FGFRs in semisolid media. Stable transfectants of NIH3T3 cells expressing the indicated receptors were suspended in DMEM containing 0.4% Seaplaque-agarose and 10% calf serum. Photographs were taken using a light microscope after 2 weeks. Results are representative of at least two independent experiments

FGFR2 activation by C-terminal alterations MV Lorenzi et al

domain. In contrast, a receptor carrying the smallest deletion in the C-terminal domain, D802-, showed slightly lower transforming activity (0.26103 f.f.u./ pmol DNA). These results suggest that the C-terminal domain of FGFR2 contains determinants which normally suppress the activity of the receptor in NIH3T3 ®broblasts. The ability of NIH3T3 transfectants expressing FGFR2 derivatives to grow in semisolid medium was also examined. NIH3T3 cells transfected with FGFR2WT or vector alone did not induce colony formation eciently in soft agar (Figure 1b). Cells expressing KGFR formed colonies eciently but the relative activity of the receptor in this assay was lower compared to its transforming activity (see Figure 2a) suggesting that autocrine stimulation by KGF is less ecient for supporting anchorage-independent growth. The cloning eciency of KGFR was approximately tenfold less than that of FGFR2-ROS, another constitutively active FGFR2 (data not shown). Similar to their activity in focus forming assays, all the receptors containing deletions in their C-terminal domains induced colony formation in soft agar. Cells expressing D8027, the receptor which showed the lowest activity in focus forming assays, also displayed the lowest eciency for colony formation in soft agar.

The activity of two other C-terminal deletion mutants, D7647 and D7897, was comparable to that observed for cells expressing FGFR2-ET while D7767 transfectants displayed slightly higher activity in this assay. Taken together, these ®ndings indicate that deletion of the C-terminal domain of FGFR2 can activate the receptor in the absence of ligand and suggest that the C-terminus of FGFR2 is dispensable for both focus formation and anchorage-independent growth. Tyrosine residues in the C-terminus of FGFR2 regulate receptor activity The C-terminal domain of FGFR2 contains tyrosine residues at positions 770, 780, 784, 806 and 813. Carboxyl terminal tyrosines in the tyrosine kinases c-Src and insulin receptor have been shown to negatively regulate the enzymatic activity of the proteins (see Hunter, 1987; Pang et al, 1994). Therefore, we postulated that the loss of tyrosine residues in the C-terminal domain of the receptor may be responsible for the higher transforming eciency of the mutant FGFRs. The sequence surrounding tyrosine 770 in FGFR2 is a potential binding site for the SH2 domain of phospholipase-C-g. Mutation of the corresponding tyrosine in FGFR1 (tyrosine 766) had

a

b

FGFR2-WT

F813

F806

F784

F780

F780/F813

Figure 2 Transforming activity and soft agar growth of FGFRs containing C-terminal tyrosine to phenylalanine substitutions. The transforming activity (a) and growth in semisolid media (b) of FGF receptors with C-terminal tyrosine to phenylalanine substitutions at positions 813 (F813), 806 (F806), 784 (F784), 780 (F780) or at both 813 and 780 (F780/F813) were performed as described in Figure 1. The location of substituted phenylalanine residues within the C-terminus is indicated by an F within the open box

819

FGFR2 activation by C-terminal alterations MV Lorenzi et al

no e€ect on mitogenic activity, cell proliferation, or cell cycle progression (Peters et al., 1992; Mohammadi et al., 1992; Huang et al., 1995) of cells expressing this receptor. Further, C-terminal deletion mutants of FGFR2 (D7767 and D7647) ¯anking this site displayed similar transforming activities (see Figure 1) suggesting that this tyrosine did not play a signi®cant role in regulating the transforming activity of the receptor. To examine the role of the other tyrosine residues in receptor activity, each tyrosine residue at positions 780, 784, 806 or 813 in the C-terminus of the wild type FGFR2 cDNA was mutated to phenylalanine and these constructs were tested for their ability to transform NIH3T3 cells. The transforming activity of three of these mutant receptors (F813, F780 and F784) was equivalent to that of FGFR2-ET (Figure 2a), with transforming activities in the range of 102 focus forming units per pmol DNA, but approximately 10fold less active than that of KGFR or PDGF-BB (data not shown). These data suggested that tyrosine residues negatively regulate receptor activity. However, the introduction of a double mutation (F780/F813) into these two regions of FGFR2 did not exhibit synergistic or even an additive e€ect on the transforming activity. Taken together, C-terminal alterations can upregulate receptor activity, but not enough to completely override the negative e€ect of the ligand binding domain. In contrast, transfection of FGFR2 cDNAs with phenylalanine substitutions at positions 806 did not mediate transformation of NIH3T3 cells (Figure 2a), suggesting that this tyrosine residue is dispensable for receptor activity when the C-terminal regulatory domain is present. Cells expressing receptors with tyrosine substitutions at positions 813, 784, and 780 showed similar abilities to form colonies in soft agar (Figure 2b). The receptor carrying a double substitution (F813/F780) showed a

a

comparable cloning eciency as the receptors carrying either single mutation (F813 or F780) alone. In contrast, cells expressing FGFR2-WT did not show signi®cant activity in this assay. Transfectants expressing the F806 receptor did exhibit a slight elevation in soft agar growth compared to FGFR2-WT expressing cells but less than F813, F784, F780 or F813/F780 cells suggesting that F806 may be on the verge of transformation. Nonetheless, the ability of FGFR2 constructs with phenylalanine substitutions at tyrosines 813, 784 and 780 to induce transforming activity and colony formation indicate that these tyrosines exert a negative e€ect on receptor activity and suggest that the loss of these tyrosines may underlie the activity of FGFR2 variants with C-terminal deletions. The higher transforming activity of receptors with C-terminal alterations is not accompanied with overall increases in receptor phosphorylation or MAPK activation To assess the e€ects of the di€erent C-terminal mutations on receptor phosphorylation, FGFR2 was precipitated from the cell lysates prepared from stable transfectants of the receptor constructs with an antiFGFR2 antibody (aFGFR2) and analysed for phosphorylation levels by immunoblot analysis with anti-phosphotyrosine (aPTYR) antibodies. The expression levels of the di€erent FGFR2 constructs in NIH3T3 transfectants were roughly equivalent with the exception of D7767 and D7647 which showed slightly lower expression levels (Figure 3a, lower panel). The receptor in the FGFR2 transfectants was identi®ed as a discrete band of around 100 kD and a broad band at 180 kD, while the receptor in the KGFR transfectants was identi®ed as 110 kD and 90 kD proteins. The slower migrating band of each

FGF:

KGFR

∆764-

∆776-

∆789-

∆802-

F780/F813

F784

F780

F806

F813

FGFR2-ET

FGFR2-WT

b pCEV29

820

αp42/p44-PTyr204

–+ –+ –+ –+ –+ –+ –+ –+ –+ –+ –+ –+ –+

200αPTyr 97200αFGFR2 97-

Figure 3 Activation of receptor phosphorylation and MAPKs in transfectants of mutated FGF receptors. (a) Expression levels and phosphorylation of FGFRs in NIH3T3 transfectants. One milligram of total cellular lysate from the indicated cells was immunoprecipitated with an anti-FGFR2 antiserum (aFGFR2) and immunoblotted with antiphosphotyrosine antibodies (aPTyr, upper blot). Membranes were subsequently stripped and reprobed with aFGFR2 (lower panel). Where indicated by a (+), cells were stimulated with basic FGF (50 ng/ml) or in the case of KGFR transfectants, KGF (50 ng/ml) for 15 min prior to cell harvesting. The location of molecular weight standards is shown on the left of the ®gure. (b) Activation of MAPKs in transformed NIH3T3 cells. Fifty micrograms of cellular lysate was prepared as described above and resolved on a 8 ± 16% gradient gel. Immunoblots were probed ®rst with antibodies speci®c for the activated forms of p42/ERK2 and p44/ERK1 MAPKs phosphorylated on tyrosine 204 (ap42/p44-PTyr2-4, inset). Blots were then stripped and reprobed with an anti-p42/p44 antisera to determine the total amount of these MAPKs in the lysates. ERK activity was expressed as the ratio of the phosphorylated to non-phosphorylated p42/ERK2 and p44/ERK1 and is depicted graphically in (b). ERK activity (n=2) was determined in cells which were serum-starved (black-®lled columns) or bFGF or KGF treated (gray-®lled columns) as described in (a)

FGFR2 activation by C-terminal alterations MV Lorenzi et al

receptor represents post-translationally modi®ed forms of FGFR2, since this band could be cross-linked by 125 I-acidic FGF and treatment with enzymes which speci®cally degrade glycosaminoglycans reduced the molecular size of the FGFRs comparable to the predicted size (Sakaguchi et al., submitted for publication). In addition to the 180 kD band, a lower molecular weight protein of 100 kD was identi®ed in transfectants expressing FGFR2-WT or receptors carrying tyrosine to phenylalanine substitutions. In transfectants expressing FGFR2-ET and C-terminal derivatives of the receptor, the size of this protein was smaller. This lower band in the transfectants may represent unmodi®ed receptor since the size of this protein roughly correlates with the size predicted from the various expression constructs. Interestingly, transfectants of receptors F813, F813/F780, or D802 displayed a much higher level of the lower molecular weight species suggesting that the loss of tyrosine 813 e€ects the post-translational processing of the receptor to its mature form. Immunoblot analysis of the same immunoprecipitates with aPTYR antibodies demonstrated that only the KGFR, which was stimulated by KGF secreted by NIH3T3 cells, was highly phosphorylated on tyrosine even without the addition of exogenous KGF (Figure 3a, upper panel). In contrast, FGFR2-WT displayed a low basal level of receptor phosphorylation. The basal phosphorylation of FGFRs with tyrosine to phenylalanine substitutions and three of the C-terminal derivatives of FGFR2 (FGFR2-ET, D8027 and D7897) was equivalent to that of the wild type receptor. Tyrosine phosphorylation of FGFRs with the C-terminal deletions of D7647 and D7797 was much lower than that of FGFR2-WT. After stimulation with bFGF, all the receptors showed a similar increase in phosphorylation, with the exception of D7647 and D7767, indicating that most of the receptors were capable of undergoing ligand-induced receptor phosphorylation. The absence of signi®cant receptor phosphorylation following bFGF treatment in D7647 and D7767 transfectants suggested the sequences missing in these receptors are important for receptor autophosphorylation. In addition, D7647 and D7767 were expressed at much lower levels compared to other FGFR isoforms (Figure 3a) suggesting that the loss of C-terminal sequences in these receptors may also a€ect protein stability. Nonetheless, despite lower protein and tyrosine phosphorylation levels the D7647 and D7767 receptors still displayed high transforming activities (see Figure 1). Taken together, these results indicated that the increased transforming activity and elevated colony formation in soft agar of receptors with C-terminal alterations does not correlate with an overall increase in receptor phosphorylation. The lack of elevated receptor phosphorylation in the di€erent transformed NIH3T3 cells expressing mutant FGFRs prompted us to examine signaling events downstream of receptor autophosphorylation. One downstream event transduced by FGF receptors in response to ligand is the activation of the MAP kinases p42/ERK2 and p44/ERK1 (Shaoul et al., 1995; Wang et al., 1994). We therefore evaluated the activation status of MAPKs in the NIH3T3 cell stable transfectants using an antiserum speci®c for the

phosphorylated forms of p42/ERK2 and p44/ERK1. As shown in Figure 3b, stimulation of vector or FGFR2-WT transfectants with bFGF resulted in an approximate three- to fourfold induction in the tyrosine-phosphorylated forms of p42/ERK2 and p44/ ERK1. In contrast, the MAP kinase activity in cells expressing mutant FGFRs was not increased compared to that of unstimulated cells expressing the wild type receptor. Interestingly, while the expression of constitutively active H-Ras did result in an increase in MAP kinase phosphorylation, expression of KGFR or FGFR2-ROS, and FGFR2 isoform activated by KGF in NIH3T3 cells and a constitutively active form of FGFR2, respectively, did not signi®cantly induce MAP kinase phosphorylation (Figure 3b). However, cells expressing either KGFR or FGFR2-ROS, did show an increase in the phosphorylation of MAP kinases in response to KGF or bFGF, respectively, indicating that the pathway is functional in these cells. These results indicated that, unlike H-Ras transformed NIH3T3 cells, cells expressing mutant or constitutively active FGFR2 isoforms do not require sustained MAP kinase activation to elicit cellular transformation. Expression of FGFRs in human tumor cell lines The overexpression or alteration of receptor tyrosine kinases has been implicated in the development of several types of human malignancies (see Aaronson, 1992). K-SAM, an isoform of FGFR2 containing a Cterminal truncation, was identi®ed as an ampli®ed gene in a gastric carcinoma cell line KATO-III (Hattori et al, 1990). Recently, K-SAM-IIC3, an alternatively spliced isoform encoding a C-terminal truncation, was found to be preferentially expressed in gastric carcinoma cells including KATO-III (Itoh et al., 1994). Our ®ndings suggested that overexpression of FGFR2 isoforms with truncated C-terminal domains could impart enhanced growth properties without an overall increase in receptor phosphorylation. To examine this possibility in more detail, we ®rst screened RNA isolated from several human tumor cell lines by Northern analysis to identify cells lines that overexpress FGFR2 mRNA. For this analysis we utilized a DNA hybridization probe encoding a portion of the extracellular domain common to both the FGFR2/Bek and KGFR receptors. In RNA isolated from cell lines derived from gastric and colon carcinomas, FGFR2/KGFR mRNA was expressed at very high levels in KATO-III and SNU-16 (Figure 4a), consistent with the ampli®cation of the FGFR2 gene in these tumor cells (Hattori et al., 1990; Mor et al., 1993). Other gastric and colon carcinoma cells, with the exception of HA117, showed expression of FGFR2/KGFR similar to that of B5/589, a normal human mammary epithelial cell line which expresses KGFR. Analysis of other tumor cell lines derived from kidney, bladder, breast, lung or liver showed increased expression of FGFR2/KGFR mRNA compared to B5/ 589. These included A704 (kidney, Figure 4b), MDAMB-415 and ZR-75-1 (breast, Figure 4c), and HA146 (lung, Figure 4d). The overexpression of FGFR2/ KGFR in these cells as well as in KATO-III and SNU16 suggests that the overexpression of FGFR2/KGFR mRNA may be selected for in the development of these tumors.

821

FGFR2 activation by C-terminal alterations MV Lorenzi et al

HA617

HA2008

N87

AGS

294621/2193

HA1233

HCT116

HT29

SW1116

A498

A704

HA142

HA212

HA251

HA383

HA442

CAKI-1

CAKI-2

MDA-MB-134IV

MDA-MB-175

MDA-MB-231

MDA-MB-361

MDA-MB-415

MDA-MB-453

MDA-MB-465

MDA-MB-468

SK-BRI-II

SK-BRI-III

HA146

HA188

MCF-7A

BT474 A549

13741

SW1271

HA1182

d

SK-HEP-1

BT20(SK) A427

MCF-7

B5/589 B5/589

ZR-75-1

M426

c

19795/2019

HA117

HA114

5637

EJTR-2BCL2

J82

T24

HT1376

HA1054

b

BT483

SNU-16

KATO-III HA698

SNU-5

OKAJIMA HA604

SNU-1

B5/589 B5/589

WRL 68

M426 M426

a

20471/2202

indicating a strong correlation between mRNA and protein expression. The lower panel of Figure 5 shows phosphotyrosine levels of the receptor. Despite the extremely high expression of the receptor in KATO-III cells, its phosphorylation level was relatively low, indicating that FGFR2 activity in KATO-III cells is mainly due to its overproduction. In contrast, higher levels of phosphotyrosine signals were detected in three tumor cell lines, MCF7 (breast), HA1213 (pancreas) and HA146 (lung), suggesting that FGFR2 has been activated in these cell lines by some mechanism(s) including autocrine stimulation by ligands secreted by these cells. While the expression levels of FGFR in HA1213 and HA146 were relatively high (Figure 5,

We next examined the expression and phosphorylation state of FGFR2/KGFR protein in cells overexpressing receptor mRNA, such as KATO-III and HA146, as well as in other tumor cell lines derived from tumors distinct from those in Figure 4. As shown in Figure 5 upper panel, similar to our observations at the mRNA level, a large amount of FGFR2/KGFR protein was expressed in KATO-III cells. Five other tumor cell lines, A673 (rhabdomyosarcoma), HA146 (lung carcinoma), HA1213 (pancreas carcinoma), HA1780 (ovarian carcinoma), and MDA-MB-415 (mammary carcinoma), also expressed the receptor at relatively high levels. Both MDA-MB-415 and HA 146 cells also expressed high levels of receptor mRNA

M426

822

Figure 4 Expression of FGFR2/KGFR mRNA in human tumor cell lines. Northern blots of total RNA isolated from tumor cell lines derived from di€erent human tissues were hybridized to a FGFR2 cDNA probe encoding a portion of the extracellular domain common to both FGFR2 and KGFR. RNA from two non-tumor cells lines, an embryonic ®broblastic cell line (M426) and a mammary epithelial cell line (B5/589) was also included in this analysis. FGFR2/KGFR mRNA was analysed in tumor cell lines derived from (A) gastric (Okajima, KATO-III, WRL 68, SNU-1, SNU-5, SNU-116, HA114, HA117, HA167, HA2008, N87, AGS, 29462/2193) or colon carcinomas (HA1233, HCT116, HT29, SW1116, 20471/2202, 19795/2019), (b) bladder (HA604, HA698, HA1054, HT1376 T24, J82, EJTR-2BCL2, 5637) or kidney carcinomas (A498, A704, HA142, HA212, HA251, HA383, HA442, CAKI-1, CAKI-2). (c) mammary carcinomas (BT20(SK), BT474, SK-BRI-II, SK-BR-III, ZR-75-1, MCF-7, MCF-7A, BT483, MDA-MB-175, MDA-MB-231, MDA-MD-361, MDA-MB-415, MDA-MB-453, MDA-MB-465, MDA-MB-468), and (d) lung (A427, A549, HA146, HA188, HA1182, SW 1271) or liver carcinomas (13741, SK-HEP-1). FGFR2/KGFR mRNA was detected as a 4.5 kb mRNA in all panels, with the exception of KATO-III (a) and MDA-MB-415 (c) which expressed smaller transcripts. The equal amounts of RNA loaded and transferred was visualized by ethidium bromide staining of RNA gels before and after transfer to membranes

FGFR2 activation by C-terminal alterations MV Lorenzi et al

A673

HA146

HA1213

A875

HA442

HA1780

MB415

MCF7

HA1984

SK-ES-1

HOS

PC3BM

PC3

HA690

HA285

HA153

KATOIII

M426

823

200 -

αFGFR2

97 -

200 -

αPTyr

97 -

Figure 5 Expression of phosphorylation of FGFR2 protein in cell lines derived from human tumors. Soluble lysates from cell lines derived from human tumors were analysed as described in Figure 3. The tumours from which the cell lines were derived are designated as follows: M426 (embryonic lung ®broblast), KATOIII (gastric), HA153 (neuroblastoma), HA285 (astrocytoma), HA690 (astrocytoma), PC3 (prostate), PC3BM (prostate, bone metastasis), HOS (osteosarcoma), SK-ES-1 (Ewing sarcoma), HA1984 (Ewing sarcoma), MCF7 (mammary), MB415 (mammary), HA1780 (ovarian), HA442 (kidney), A875 (melanoma), HA1213 (pancreas), HA146 (lung), and A673 (rhabdosarcoma). The tumor cell line MDA-MB415 was abbreviated MB415 in the ®gure

upper panel), MCF7 expressed very low levels of both receptor protein and mRNA (see Figure 4c) indicating that FGFR2 is highly phosphorylated and therefore likely to be highly activated in these cells. We screened a MCF7 cDNA library for FGFR2/KGFR and isolated cDNAs encoding KGFR isoforms with normal C-terminal domains suggesting that the receptor may be activated by an autocrine mechanism in these cells (data not shown). Nonetheless, these data suggested that both overexpression of FGFR2/KGFR isoforms with truncated C-termini, as in KATO-III cells, or the higher phosphorylation of FGFR2/KGFR isoforms, as in MCF-7 cells, may have contributed to the neoplastic progression of these cell lines. Discussion In the present study we established that expression of FGFR2 variants with C-terminal alterations can impart enhanced growth properties to NIH3T3 cells in a ligand-independent manner. While overexpression of the wild type receptor did not show signi®cant transforming activity or growth in semisolid medium, receptors containing deletions in the C-terminal domain induced morphological transformation and anchorage-independent growth of NIH3T3 cells. Deletions in the C-terminal domain of two other receptors, CSF-1 receptor and erbB2, have also been shown to activate these receptors with respect to transforming activity (Woolford et al., 1988; Akiyama

et al., 1991). A common feature of both of these receptors is that mutation of C-terminal tyrosine residues, which were lost in the receptors with C-terminal deletions, in receptors with full-length C-termini had no e€ect on transforming activity (Roussell et al., 1987; Akiyama et al., 1991). In contrast, FGFRs engineered to encode tyrosine to phenylalanine substitutions at positions 813, 784 or 780 in the C-terminus of FGFR2 showed ligand-independent activation similar to that of FGFRs with Cterminal truncations. Thus, the carboxyl terminus of FGFR2 can negatively regulate the transforming activity of the receptor in a manner dependent, in part, on C-terminal tyrosine residues. The increased transforming activity of FGFR2 variants with carboxyl terminal alterations was not associated with an overall increase in the phosphorylation state of the receptor. Tyrosine 770 in the Cterminus of FGFR2 is conserved in the other FGFRs and corresponds to tyrosine 766 in FGFR1 which has been shown to be a major site of receptor autophosphorylation (Mohammadi et al, 1991; Hou et al., 1993) and when phosphorylated, the binding site for the SH2 domain of phospholipase C-g-1 (Mohammadi et al., 1991). However, it is unlikely that the reduced receptor phosphorylation and increased transforming activity observed with FGFR2 C-terminal derivatives in this study can be attributed to the loss of tyrosine 770 in FGFR2 since only one of the mutants described here (D7647) lacked this tyrosine. The reduced phosphorylation state of these mutant receptors could

FGFR2 activation by C-terminal alterations MV Lorenzi et al

824

re¯ect the loss of other C-terminal phosphorylation sites since it is not known which C-terminal residues of FGFR2 are phosphorylated in vivo. Interestingly only two of the C-terminally altered receptors described here, D7647 and D7767, failed to exhibit an overall increase in tyrosine phosphorylation in response to bFGF stimulation. It is possible that the lack of FGFresponsive phosphorylation in D7647 and D7767 re¯ects loss or perturbation in phosphorylation of Y770, respectively. Alternatively, residues between 776 and 789, such as tyrosines 780 and 784, may be important in receptor phosphorylation. However, unlike D7647 and D7767, receptors carrying mutations at tyrosines 780 or 784 still displayed overall increases in receptor phosphorylation in response to bFGF suggesting that the phosphorylation of other sites on these receptors may be elevated to compensate for the loss of these tyrosines. Nonetheless, unlike constitutively activated receptors, such as KGFR, none of the transforming C-terminal derivatives of FGFR2 increased the basal level of receptor phosphorylation suggesting that tyrosines 780, 784, and 813 in the C-terminus of FGFR2 may repress transformation by a mechanism not involving receptor phosphorylation. Our results indicate that deletion of the FGFR2 C-terminus by alternative splicing can increase the transforming activity of the receptor without increasing the basal kinase activity. It has been proposed that the C-terminal domain of tyrosine kinases such as EGFR, insulin receptor, and c-Src can act as a competitive `pseudo-substrate' for the kinase thereby modulating activity (Hunter, 1987; Bertics and Gill, 1985; Kaliman et al., 1993). In this regard it is interesting to note that the sequence surrounding tyrosine 813 (PQYPH) in FGFR2 is similar to the sequence ¯anking Y527 (PQYQP) in the C-terminus of c-Src (see Hunter, 1987). Phosphorylation of tyrosine 527 in c-Src has been shown to induce the binding of this phosphotyrosine (and ¯anking sequence) to the SH2 domain of the protein to inhibit kinase activity (Superti-Furga et al., 1993). Further, mutation of this residue (Y to F) removes the inhibitory e€ect and dramatically increases the transforming activity and kinase activity of Src (Piwnica-Worms et al., 1987). The corresponding mutation in FGFR2 (F813) reported in this study also increased the transforming activity of the receptor but without an associated increase in receptor phosphorylation. Tyrosine 527 of c-Src has been proposed to act as a regulator of protein function by stabilizing a repressed protein conformation through an allosteric mechanism (Cooper and Howell, 1993). Tyrosine 813 and/or other tyrosines in the C-terminus of FGFR2 may also be utilizing a similar mechanism to modulate FGFR2 activity. Alternatively, the negative regulation of receptor activity we observe with receptors containing tyrosine to phenylalanine substitutions in the C-terminus may re¯ect an overall disruption in the structural integrity of this domain. In this light it will be of interest to determine if mutation of other residues besides tyrosine in the C-terminal domain of FGFR2 can also induce the transforming activity of the receptor. Growth factor receptors are frequently overexpressed or undergo structural rearrangements in tumor cells (Aaronson, 1992; Sawyers and Denny, 1994). Tumor-speci®c chromosomal rearrangements involving

growth factor receptors frequently convert nontransforming receptors into potent oncogenes through the loss and replacement of the ligand-binding domain which results in the constitutive activation of the receptor's enzymatic activity (Sawyers and Denny, 1994). On the other hand, overexpression of certain receptor tyrosine kinases, such as erbB2, is able to induce ligand-independent cellular transformation while that of others, such as epidermal growth factor receptor, is not sucient to induce transformation suggesting that the presence of a receptor-speci®c structural motif(s) or other factors are necessary to override the negative e€ect of the ligand-binding domain (Di Fiore et al., 1987a,b). One such motif appears to be the C-terminal domain of the receptor since our results indicate that overexpression of FGFR2 is unable to induce transformation of NIH3T3 cells whereas overexpression of C-terminally altered receptors can confer ligand-independent transformation of these cells but without increasing receptor phosphorylation. Consistent with these ®ndings we found that K-SAM, a C-terminally truncated KGFR, is expressed at a very high level in the gastric carcinoma cell line KATO-III, but is poorly phosphorylated. Our results indicate that alterations in the C-terminus of the receptor coupled with receptor overexpression can overcome the negative e€ect of the ligand-binding domain but that such alterations are unable to increase receptor autophosphorylation. It is interesting to note that the C-terminal domain of the receptor is a region less conserved among FGF receptor family members, suggesting that this domain may impart a di€erent function in the other FGF receptors. In this regard it will be of considerable interest to further clarify the role of FGFRs overexpression in human malignancy. Materials and methods Plasmid DNAs The FGFR2-WT, FGFR2-ROS and FGFR2-ET cDNAs used in these studies have been described previously (Lorenzi et al., 1996). All three cDNAs encode receptors with two immunoglobulin (Ig)-like domains and an acidic region. The KGFR cDNA used in these studies has been described elsewhere (Miki et al., 1992). The H-RASV12 cDNA was the kind gift of Dr J Silvio Gutkind. All cDNAs were cloned into the expression vector pCEV29 Site-directed mutagenesis Site-directed mutagenesis was performed on FGFR2-WT using a two-step polymerase chain reaction (PCR) method (Gak et al., 1992). C-terminal tyrosine codons in FGFR2WT cDNA were modi®ed using oligonucleotides encoding tyrosine to phenylalanine substitutions (bold) at positions 813 (5'-TGCCTCAGTTTCCACACATATA-3'), 806 (5'-CC ATGCCTTTTGAACCCTGT-3'), 784 (5'-TTCTCCTAGT TTCCCCGACA-3'), or 780 (5'-CTCGAACAGTTTTCTC CTAG-3') of the FGFR2. Amino acid and nucleotide numbering are according to a three immunoglobulin domain form of human FGFR-2 as presented in Ron et al. (1993). PCR also was used to construct FGFR2 with partial deletions of the C-terminus. The construction of the FGFR2 truncated at position 764 has been described elsewhere (Lorenzi et al., 1996). For the remaining Cterminal derivatives of FGFR2, termination codons (bold)

FGFR2 activation by C-terminal alterations MV Lorenzi et al

were inserted by PCR to introduce deletions of the Cterminus at positions 776 (5'-GCAAAGCTTCACTGGGTGAGATCCAGTATTCCTC-3'), 789 (5'-GCAAAGCTTCATGTGTCGGGGTAACTAGGAGAATA-3') and 802 (5'-GCAAAGCTTCATCCAGA AAACA CAGAATCGTCCCC-3') of the FGFR2-WT. A HindIII site (underlined) was added to each of the termination primers to facilitate ligation with FGFR2-WT. PCR products encoding Cterminal deletions and point mutations of FGFR2 were digested with BglII and HindIII to excise a 0.38 kb fragment containing the mutated sequence(s). This fragment was ligated into the FGFR2-WT cDNA which had the corresponding BglII ± HindIII fragment removed. The sequence between these two restriction sites was con®rmed in all mutants by using the Sequenase kit 2.0 (US Biochemical). Transfection analysis and soft agar assays Expression constructs (0.01 ± 1.0 mg/plate) were transfected into NIH3T3 ®broblasts by the calcium-phosphate precipitation as described (Wigler et al., 1978). Transformed foci were counted after approximately three weeks in culture. Mass cultures of NIH3T3 cells stably expressing the recombinant proteins were obtained by selection with G418 (750 mg/ml). To test anchorage-independent growth, cell suspensions were plated at 10-fold serial dilutions in semisolid agarose medium containing Dulbecco's modi®ed Eagle medium (DMEM) supplemented with 10% calf serum and 0.45% seaplaque agarose (FMC). Cell suspensions were given fresh media every 3 days. Visible colonies were scored and photographed after 14 days. Northern analysis Total RNA was isolated from human tumor cell lines using RNAzol (Teltest) following the manufacturer's instructions. Northern blots containing 20 mg of total RNA extracted from human tumor cell lines were prepared and processed as described (Lorenzi et al., 1995). A 666 bp ApaI fragment encoding amino acids 82 ± 304 of the human KGFR cDNA was used a hybridization probe. All RNA samples were assessed by ethidium bromide staining of gels

prior to and after transfer to nylon membranes to ensure that an equal quantity of RNA was transferred. Protein analysis NIH3T3 cells stably transfected with the indicated expression constructs grown in the presence of G418 or human tumor cell lines were washed once in phosphate bu€ered saline (PBS) and incubated at 378C overnight in serum-free DMEM. Cells were stimulated with basic FGF or KGF at 50 ng/ml for 15 min where indicated. Cells were lysed in 50 mM HEPES (pH 7.5), 1% Triton X-100, 50 m M NaF, 10 mM sodium pyrophosphate, 1 mM Na3VO4, 5 mM EDTA, 1 mM phenylmethylsulfonyl ¯uoride, 10 mg/ml aprotinin, and 10 mg/ml leupeptin. Lysates (1 mg) were immunoprecipitated with a rabbit polyclonal antiserum generated against the amino acids 476 ± 822 of the cytoplasmic domain of FGFR2 (anti-FGFR2, 1:100 dilution). Immunocomplexes were collected by incubation with Gamma Bind G Plus (Pharmacia LKB). Immunoprecipitated proteins were resolved on SDS ± PAGE (7.5%) and immunoblotted with anti-phosphotyrosine (1:1000, Upstate Biotechnology, Inc.). Bound antibody was visualized using [125I]protein A. Following removal of anti-phosphotyrosine antibody with 0.5 M glycine (pH=2.0) and 0.05% Tween-20, membranes were reprobed with an anti-FGFR2 antisera (1:500). For analysis of p42/ERK2 and p44/ERK1 activation, lysates of NIH3T3 transfectants were prepared as described above. Fifty micrograms of each lysate was resolved on a 8 ± 16% polyacrylamide gel and immunoblotted with an antisera speci®c for the tyrosine-204 phosphorylated forms of p42 and p44 MAP kinases (ap42/p44-PTYR204, New England Biolabs, NEB). Following removal of ap42/p44-PTYR204, blots were reprobed with an anti-p42/p44 antisera. For these experiments, bound antibody was visualized using the chemiluminescent ECL kit (Amersham) following the manufacturer's instructions. Acknowledgements We thank Dr JH Pierce for support and Dr AML Chan for advice on site-directed mutagenesis.

References Aaronson SA. (1992). Science, 254, 1146 ± 1153. Akiyama T, Matsuda S, Namba Y, Saito T, Toyoshima K and Yamamoto T. (1991). Mol. Cell. Biol., 11, 833 ± 842. Basilico C and Moscatelli D. (1992). Adv. Cancer Res., 59, 115 ± 165. Bertrics PJ and Gill GN. (1985). J. Biol. Chem., 260, 14642 ± 14647. Champion-Arnaud P, Ronsin C, Gilbert E, Gesnel MC, Houssaint E and Breanthnach R. (1991). Oncogene, 6, 979 ± 987. Chellaiah AT, McEwen DG, Werner S, Xu J and Ornitz DM. (1994). J. Biol. Chem., 269, 11620 ± 11627. Cooper JA and Howell B. (1993). Cell, 73, 1051 ± 1054. Di Fiore PP, Pierce JH, Kraus MH, Segatto O, King CR and Aaronson SA. (1987a). Science, 237, 178 ± 182. Di Fiore PP, Pierce JH, Fleming TP, Hazan R, Ullrich A, King CR, Schlessinger J and Aaronson SA. (1987b). Cell, 51, 1063 ± 1070. Dionne CA, Crumley G, Bellot F, Kaplow JM, Searfoss G, Ruta M, Burgess W, Jaye M and Shlessinger J. (1990). EMBO J., 9, 2685 ± 2692. Finch PW, Cunha GR, Rubin JS, Wong J and Ron D. (1995). Devlp. Dyn., 203, 223 ± 240. Gak E, Taylor WG, Chan AML and Rubin JS. (1992). FEBS Lett., 311, 17 ± 21.

Gao G and Goldfarb M. (1995). EMBO J., 14, 2183 ± 2190. Hattori Y, Odagiri H, Nakatani H, Miyagawa K, Naito K, Sakamoto H, Katoh O, Yoshida T, Sugimura T and Terada M. (1990). Proc. Natl. Acad. Sci. USA, 87, 5983 ± 5987. Hou J, McKeehan K, Kan K, Carr SA, Huddleston MJ, Crabb JW and McKeehan WL. (1993). Protein Sci., 2, 86 ± 92. Huang J, Mohammadi M, Rodrigues GA and Schlessinger J. (1995). J. Biol. Chem., 270, 5065 ± 5072. Hunter T. (1987). Cell, 49, 1 ± 4. Itoh H, Hattori Y, Sakamoto H, Ishii H, Kishi T, Sasaki H, Yoshida T, Koono M, Sugimura T and Terada M. (1994). Cancer Res., 54, 3237 ± 3241. Kaliman P, Baron V, Alengrin F, Takata Y, Webster NJG, Olefsky JM and Obberghen EV. (1993). Biochemistry, 32, 9539 ± 9544. Kan M, Wang F, Xu J, Crabb JW, Hou J and McKeehan WL. (1993). Science, 259, 1918 ± 1921. Keegan K, Johnson DE, Williams LT and Hayman MJ. (1991). Proc. Natl. Acad. Sci. USA, 88, 1095 ± 1099. Lorenzi MV, Long JE, Miki T and Aaronson SA. (1995). Oncogene, 10, 2051 ± 2055. Lorenzi MV, Horii Y, Yamanaka R, Sakaguchi K and Miki T. (1996). Proc. Natl. Acad. Sci. USA, 93, 8956 ± 8961.

825

FGFR2 activation by C-terminal alterations MV Lorenzi et al

826

Mansukhani A, Moscatelli D, Talarico D, Levytska V and Basilico C. (1990). Proc. Natl. Acad. Sci. USA, 87, 4378 ± 4382. Miki T, Fleming TP, Bottaro DP, Rubin JS, Ron D and Aaronson SA. (1991). Science, 251, 72 ± 75. Miki T, Bottaro DP, Fleming TP, Smith CL, Burgess WH, Chan AML and Aaronson SA. (1992). Proc. Natl. Acad. Sci. USA, 89, 246 ± 250. Mohammadi M, Honegger AM, Rotin D, Fischer R, Bellot F, Li W, Dionne CA, Jaye M, Rubinstein M and Schlessinger J. (1991). Mol. Cell. Biol., 11, 5068 ± 5078. Mohamadi M, Dionne CA, Li W, Li T, Spivak T, Honegger AM, Jaye M and Schlessinger J. (1992). Nature, 358, 681 ± 684. Mor O, Ranzani GN, Ravia Y, Rotman G, Gutman M, Manor A, Amadori D, Houldsworth J, Hollstein M, Schwab M and Shiloh Y. (1993). Cancer Genet. Cytogenet., 65, 111 ± 114. Pang L, Milarski KL, Ohmichi M, Takata Y, Olefsky JM and Saltiel AR. (1994). J. Biol. Chem., 269, 10604 ± 10608. Partanen J, Makela TP, Eerola E, Korhonen J, Hirvonen H, Claesson-Welsh L and Alitalo K. (1991). EMBO J., 10, 1347 ± 1354. Peters KG, Marie J, Wilson E, Ives HE, Escobedo J, Del Rosario M, Mirda D and Willaims LT. (1992). Nature, 358, 678 ± 681.

Piwnica-Worms H, Saunders KB, Roberts TM, Smith AE and Cheng SH. (1987). Cell, 10, 75 ± 82. Ron D, Reich R, Chedid M, Lengel C, Cohen OE, Chan AML, Neufeld G, Miki T and Tronick SR. (1993). J. Biol. Chem., 268, 5388 ± 5394. Roussel MF, Dull TJ, Rettenmier CW, Ralph P, Ullrich A and Sherr CJ. (1987). Nature, 325, 549 ± 552. Sawyers CL and Denny CT. (1994). Cell, 77, 171 ± 173. Shaoul E, Reich-Slotky R, Berman B and Ron D. (1995). Oncogene, 10, 1554 ± 1561. Superti-Furga G, Fumagalli S, Koegl M, Courtneidge SA and Draetta G. (1993). EMBO J., 12, 2625 ± 2634. Wang JK, Gao G and Goldfarb M. (1994). Mol. Cell. Biol., 14, 181 ± 188. Wigler M, Pellicer A, Silverstein S and Axel R. (1978). Cell, 14, 725 ± 731. Wilkie AO, Morriss-Kay GM, Jones EY and Heath JK. (1995). Curr. Biol., 5, 500 ± 7. Woolford J, McAuli€e A and Rohrschneider LR. (1988). Cell, 55, 965 ± 977. Yamaguchi F, Saya H, Bruner JM and Morrison RS. (1994). Proc. Natl. Acad. Sci. USA, 91, 484 ± 488. Yan G, McBride G and McKeehan WL. (1993). Biochem. Biophys. Res. Comm., 194, 512 ± 518. Yayon A, Zimmer Y, Guo-Hong S, Avivi A, Yarden Y and Givol D. (1992). EMBO J., 11, 1885 ± 1890.