| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cell Growth & Differentiation |
Articles |
Ruhr-Universität Bochum, Med. Fakultät, Abteilung für Virologie, D-44780 Bochum, Geb. MA 6/130, Germany [M. A., M. D., R. K., V. B., P. K.], and Imperial Cancer Research Fund Laboratories, London WC2A 3PX, United Kingdom [M. J. C., G. P.]
Abstract
Fibroblast growth factor 3 (FGF3) was originally identified as the mouse proto-oncogene Int-2, which is activated by proviral insertion in tumors induced by mouse mammary tumor virus. To facilitate the biological characterization of the ligand, we have analyzed its homologue in Xenopus laevis, XFGF3. Here we confirm that the X. laevis genome contains two distinct FGF3 alleles, neither of which is capable of encoding the NH2-terminally extended forms specified by the mouse and human FGF3 genes. Unlike the mammalian proteins, XFGF3 is efficiently secreted as a Mr 31,000 glycoprotein, gp31, which undergoes proteolytic cleavage to produce an NH2-terminally truncated product, gp27. Processing removes a segment of 18 amino acids immediately distal to the signal peptide that is not present in the mammalian homologues. By inserting an epitope-tag adjacent to the cleavage site, we show that a substantial amount of the gp27 is generated intracellularly, although processing can also occur in the extracellular matrix. Two residues are also removed from the COOH terminus. To compare the biological properties of the different forms, cDNAs were constructed that selectively give rise to the larger, gp31, or smaller, gp27, forms of XFGF3. As judged by their ability to cause morphological transformation of NIH3T3 cells, their mitogenicity on specific cell types, and their affinity for the IIIb and IIIc isoforms of Xenopus FGF receptors, gp27 has a much higher biological activity than gp31. Sequence comparison revealed an intriguing similar cleavage motif immediately downstream of the signal peptide cleavage site in the NH2-terminus of mouse and human FGF3. Analysis of secreted mutant mouse FGF3 confirmed an additional NH2-terminal processing at the corresponding sequence motif. NH2-terminal trimming of Xenopus and mammalian FGF3s may therefore be a prerequisite of optimal biological activity.
Introduction
The FGFs2 are a family of related polypeptides, typically between Mr 15,000 and Mr 35,000 in size, that have been credited with many and diverse biological functions (reviewed in Refs. 1, 2, 3, 4, 5, 6, 7 ). The prototypic factors, FGF1 and FGF2 (previously acidic and basic FGF, respectively), were discovered in a wide spectrum of cell types but have aroused most interest as inducers of mesoderm, as endothelial cell mitogens, and as angiogenic factors. This latter activity suggests a possible role in tumorigenesis, and three members of the FGF family have been directly implicated as cellular oncogenes. These are FGF3 (formerly Int-2), FGF4 (formerly Hst-1 or KFGF) and FGF8, the genes for which are known targets for proviral insertion in tumors induced by mouse mammary tumor virus (8 , 9) .
In trying to understand the role of FGF3 in mammary tumorigenesis, it became clear that the gene is not expressed in normal mammary epithelial cells nor in the majority of adult mouse tissues (10, 11, 12, 13, 14) . The principal sites of expression occur in the developing embryo, where a complex series of transcripts can be detected as early as 6.5 days (13 , 15) . Expression is maintained until parturition but occurs at highly restricted sites and times. From these patterns of expression and the biochemical properties of the FGF family, it has generally been assumed that FGF3 acts as a cell-to-cell signalling molecule. However, the function of the mouse FGF3 protein has proved difficult to address because its expression and release from cells appears to be tightly regulated at the transcriptional (16, 17, 18) , translational (19 , 20) , and posttranslational levels (21, 22, 23, 24) .
A recent study on the Xenopus homologue of FGF3 has shown that XFGF3 is more efficiently secreted and a more potent mitogen than its mouse counterpart (25) . The Xenopus ligand is also more widely expressed and can interact with a broader range of high-affinity transmembrane tyrosine kinase receptors than the mouse protein. Four high-affinity receptors genes have been identified thus far in mammals (FGFR1, FGFR2, FGFR3 and FGFR4) and analogous receptor genes have been isolated from amphibians (Ref. 7 ; reviewed in Ref. 26 ). Although these receptors share the same basic structure, comprising an extracellular ligand-binding domain with two or three immunoglobulin-like motifs, a short transmembrane domain and a highly conserved tyrosine kinase element, each receptor gene can specify a variety of different forms through differential splicing and processing. Although the function of many of these FGFR variants is not yet understood, isoforms of FGFR 1, 2, and 3 that differ in the second half of the third immunoglobulin loop (designated IIIb and IIIc) exhibit distinct ligand-binding specifities (27, 28, 29, 30, 31) . XFGF3 binds with high affinity to both isoforms of mouse and Xenopus FGFR2, whereas mouse FGF3 only interacts with the IIIb isoform of FGFR2 and weakly with the IIIb isoform of FGFR1 (32 , 33) .
This species difference could reflect the additional 18-amino acid segment in the Xenopus ligand not present in the mammalian FGF3 proteins, or the fact that XFGF3 undergoes posttranslational processing. To investigate these possibilities and the significance of NH2-terminal processing for the biological activity of XFGF3, we generated mutant versions that are resistant to proteolytic cleavage, as well as epitope tagged forms that enabled us to determine whether processing was occurring in intracellular or extracellular compartments. The nonprocessed forms of XFGF3 exhibit a much lower affinity for the Xenopus FGF receptors and have lower transforming and mitogenic activity on mouse cell lines than the mature, fully processed protein. To our knowledge, this is the first report of enhanced biological activity of an FGF as a consequence of naturally occurring, proteolytic cleavage.
Results
Two Distinct FGF3 Alleles Are Present in the
Xenopus Genome.
The mouse Fgf-3 gene has the capacity to encode two primary
translation products: a Mr 28,000 product
initiating at a conventional AUG codon and a Mr
31,000 product initiating at an upstream CUG (20
, 23)
. The
CUG codon appears to be the primary initiation start site of mouse
FGF3, producing a form of the protein that partitions between the
secretory pathway and the cell nucleus, whereas the AUG-initiated form
is directed to the secretory pathway via a typical signal peptide
(20
, 23
, 34)
. In an attempt to determine whether these
unusual features of mouse FGF3 are evolutionarily conserved, we have
isolated and characterized FGF3-related cDNA clones from Xenopus
laevis. A 290-bp genomic DNA fragment, encompassing the second
exon of mouse Fgf-3, was used to screen a 
gt10 cDNA
library prepared from Xenopus RNA at embryonic stage 17.
From a total of 3 x 105 recombinant phages hybridized
under low stringency conditions, three positive clones were recovered
after plaque purification (designated B, C, and L). When the
same library was subsequently screened with probes derived from these
cDNA clones, one additional phage was recovered (designated 17).
Although 17 and C represent the same sequence in their region of
overlap, the 5' EcoRI fragment recovered in B differed at
10 residues relative to 17. As indicated in Fig. 1a
, three of these differences change the predicted amino acid sequence.
|
17 revealed an open reading frame for a
237-amino acid product, initiating at nucleotide 304 and terminating at
1015. The presumed AUG initiation codon is in a reasonably favorable
context according to the consensus for the ribosome scanning model
(35)
but is preceded by an AUG codon in the +1 reading
frame (Fig. 1a)
. In this respect, the situation is analogous
to that in the mouse, human, and zebrafish FGF3 genes and
suggests that translation of XFGF3 is tightly regulated (19
, 24
, 25
, 36)
. However, there is no possibility that the
Xenopus gene can encode an NH2-terminally
extended form of FGF3 initiated from a CUG because there are numerous
in-frame stop codons within the 150 nucleotides upstream of the AUG
(Fig. 1a
; Ref. 25
).
The predicted XFGF3 product can be readily aligned with the sequence of
mouse and Zebrafish FGF3 (Fig. 1b)
with 81 and 82%
identity, respectively, over the central region of 170 amino acids
(residues 42 to 212), and the similarity is >90%, allowing for
substitution of chemically similar amino acids. The "bridge" region
between the two blocks of homology, which varies in different FGFs
(reviewed in Ref. 3
), is characteristic for FGF3 (Fig. 1b
, arrows). However, it is the length rather
than the sequence of this bridge region that appears to be critical,
because the mouse, Zebrafish, and Xenopus forms of FGF3 show
considerable variations in sequence across this part of the molecule
(Fig. 1b)
. The three sequences also diverge completely at
the COOH terminus, downstream of residue 212 in the Xenopus
protein (Fig. 1b)
.
The most striking difference between the Xenopus and
Zebrafish FGF3 proteins on the one hand and the mouse and human
proteins on the other is the insertion of 18 amino acids immediately
after the signal peptide cleavage site (Fig. 1b)
. The
presence of this additional domain raised the possibility that it is
encoded by a separate exon that has been lost from or never acquired by
the mammalian FGF3 genes. Alternatively, it remained possible that the
cDNAs recovered from the embryonic stage 17 library represented only
one of the two genome equivalents in pseudo-tetraploid
Xenopus cells (37
, 38)
and that a direct
equivalent of the mammalian proteins could be encoded by the other
allele. To address this issue, two oligonucleotide primers based on
sequences in the 5' untranslated region of the 17 cDNA were used in
conjunction with a reverse primer in the conserved coding region to
amplify the intervening sequences in Xenopus genomic DNA by
PCR. The resultant products, 480 and 230 bp, respectively, indicated
that the 17 and genomic DNA are colinear in this region and that the
additional 18 amino acids in Xenopus FGF3 are not encoded by
a separate exon. As further confirmation, the PCR products were cloned
into plasmid vectors, and representative clones were sequenced in each
case. Of the two 480-bp clones sequenced, one was identical to 17,
whereas the second was identical to B. The two 230-bp clones also
corresponded to B. Because one of the base changes in B (a C-to-T
transition) destroys a BamHI site just downstream of the
start of translation, an additional 10 clones were tested for the
presence of this BamHI site. Of these, two had retained the
site, whereas eight had lost the site. Because these clones were
derived from genomic DNA, they presumably reflect different alleles of
the gene.
Detection of XFGF3 in the Secretory Pathway and Extracellular
Matrix.
As reported previously, biosynthesis of XFGF3 proceeds via signal
peptide cleavage at A21, glycosylation at N83, and a second proteolytic
cleavage at R45. As a result, two glycosylated products, gp31 and gp27,
accumulate in the ECM and can be displaced into the medium by addition
of soluble heparin (25)
. Alternatively, heparin may
stabilize the growth factor in the supernatant. These findings are
recapitulated in Fig. 2a
, which compares the steady-state levels of the different XFGF3-related
proteins in cell extracts, medium, and ECM from COS-1 cells transfected
with the expression plasmid Xfgf3.1. This plasmid, in which expression
is driven by the SV40 early promoter, contains the full coding domain
of XFGF3 optimized for efficient translation as described previously
(25)
. Xfgf3.1 was introduced into COS-1 cells by
electroporation, and total cell extracts, ECM, and conditioned medium
were fractionated by SDS-PAGE, transferred onto nitrocellulose
membranes, and probed with the MSD-1 monoclonal antibody (see Fig. 1b
for the epitope recognized by this antibody). Note that
both forms of the protein are secreted into the ECM and can be released
into the medium upon the addition of soluble heparin and that most of
the smaller cell-associated protein is displaced by heparin, suggesting
that it is already bound on the cell surface.
|
X3 raised
against the COOH terminal peptide of XFGF3, the staining pattern was
predominantly reticular with a superimposed juxtanuclear staining,
typical of a secreted protein (39)
, with most of the
signal concentrated in the endoplasmic reticulum. A similar staining
pattern was obtained with the MSD-1 antibody, using confocal microscopy
(Fig. 2b)
. Significantly, the confocal images were clearly
different from those observed with the secreted form of mouse FGF3
(expressed from the pKC3.2 plasmid), which has been shown previously to
accumulate in the Golgi complex of transfected COS-1 cells (Fig. 2b
; Ref. 22
).
NH2- and COOH-Terminal Processing of XFGF3.
We established previously that the smaller form of XFGF3, gp27, is
generated by cleavage at R45 that occurs within an RQRR motif, typical
of processing sites in protein precursors (40)
. To enable
us to detect any further processing, for example at the COOH terminus,
and to identify where in the cell gp27 is generated, we constructed a
version of the XFGF3 cDNA, designated Xfgf3.6, in which a FLAG epitope
(DYKDDDDK) was introduced immediately distal to the cleavage site (Fig. 3)
. This strategy has been successfully applied to other processed proteins
(41)
. Cleavage at the RQRR motif will create an
NH2-terminal FLAG epitope. The monoclonal antibody M1
selectively recognizes the FLAG tag as a free NH2-terminal
epitope and will therefore specifically detect tagged XFGF3 that has
been cleaved at R45. In addition, a sequence encoding an epitope of
human MYC was appended to the COOH terminus, allowing detection of the
XFGF3 cleavage products with the monoclonal antibody 9E10
(42)
. COS-1 cells were transiently transfected with
Xfgf3.6 or the pKC3 vector control and grown in the presence of
heparin. The FLAG- and MYC-tagged XFGF3 products were analyzed by
immunoblotting and immunofluorescence microscopy.
|
X3, detected three major
products of Mr 34,000, Mr
32,500, and Mr 29,500 in the conditioned medium
from Xfgf3.6-transfected cells (Fig. 3)
. A minor
Mr 31,000 band was also faintly visible with
MSD-I, just below the major Mr 32,500 product.
Because the Mr 34,000 form was also recognized
by 9E10, but not by M1, it presumably corresponds to the glycosylated
form of XFGF3 after signal peptide cleavage (i.e., gp31 plus
the extra residues from the epitope tags). Surprisingly, the smaller
Mr 32,500 form was not detected with either 9E10
or M1, suggesting that it had lost the COOH-terminal MYC tag but had
not undergone NH2-terminal cleavage at R45. Both the minor
Mr 31,000 form and the more prominent
Mr 29,500 product were detected by the M1 FLAG
antibody, indicating that they had been NH2-terminally
processed. Because the Mr 31,000 form was also
visible with 9E10, it most likely corresponds to a product that has
been cleaved at R45 but has not undergone the COOH-terminal cleavage to
remove the MYC epitope. In line with this interpretation, the fully
processed Mr 29,500 product was not detected
with 9E10. Interestingly, another RQRR motif occurs very close to the
COOH terminus of XFGF3 (see Fig. 1b
) and would be a likely
target for the additional processing observed in this experiment.
Cleavage at R235 would remove two amino acid residues from the COOH
terminus of XFGF3, and this would explain why the gp27 form of protein
often appears as a doublet in SDS-PAGE. Although these two residues are
included in the peptide used to generate the X3 polyclonal
antiserum, the fully processed form of the protein is still recognized
by this antiserum (Fig. 3)
.
Processing of XFGF3 Occurs in the Secretory Pathway and
Extracellular Space.
The distribution of gp27 among cells, ECM, and medium (see Fig. 2a
) suggested that it may be generated primarily in the
extracellular space. However, data obtained with the epitope-tagged
versions of XFGF3 indicated that at least some of the mature form could
be detected in the cell extract (not shown). To assess the steady-state
distribution of the various processed forms of XFGF3, COS-1 cells were
transfected with Xfgf3.6 and grown in the presence of heparin to
displace protein from the cell surface. The cells were then fixed and
permeabilized for immunofluorescence microscopy (Fig. 4,a and b)
. Under these conditions, the staining
patterns obtained with M1 and X3 were quite similar, showing
prominent reticular and paranuclear staining. The specificity of the
signal was confirmed using cells transfected with the empty vector (not
shown). In the presence of heparin to displace protein from the cell
surface, the lack of immunofluorescence with M1 or X3 if the cells
were not permeabilized prior to staining indicates that the substantial
signal detected with the M1 antibody after permeabilization of the
cells was in fact attributable to intracellular localized (Fig. 4, c and d)
. Thus, the staining pattern of the
processed forms of XFGF3, cleaved at R45, is consistent with their
localization in the Golgi compartment.
|
), displacement with
soluble heparin causes a dramatic increase in the proportion of the
mature form (gp27), suggesting that processing of XFGF3 can occur also
in the extracellular compartment.
Processing of XFGF3 by Plasmin.
Cleavage at the RQRR motif would be consistent with processing by
members of the PC family, some of which are resident in the
trans-Golgi network, or by extracellular trypsin-like
proteases such as plasmin. To test this latter possibility, we asked
whether exogenous plasminogen (as a source of active plasmin) could
promote the processing of XFGF3 in conditioned medium.
Xfgf3.1-transfected cell extracts, which were used as a source of
nonprocessed gp31, and cell supernatants, as a source of gp27, were
mixed with conditioned medium containing 0.1% FCS and 5 µg/ml
heparin, from COS-1 cells transfected with the empty vector. Increasing
concentrations of plasminogen were then added, and after a 60-min
incubation, the protein samples were analyzed by SDS-PAGE and
immunoblotting. As shown in Fig. 5
, addition of 0.1 unit/ml of plasminogen was enough to convert gp31
protein into gp27. No residual gp31 persisted at higher concentrations
of plasminogen. Because the COOH terminus of XFGF3 contains several
potential plasmin cleavage sites within and NH2-terminal of
the X3 recognition motif (see Fig. 1b
), cleavage at such
sites would result in loss of the immunoreactivity of the truncated
forms. However, under these conditions, the secreted gp27 was
remarkably resistant to plasminogen digestion (Fig. 5)
. These data
confirmed that gp27 is a derivative of gp31 and is the more stable form
in tissue culture fluid.
|
. To assess the influence of the sequence between the signal peptide and
the second cleavage motif, a deletion mutant of this region was
introduced into the Xfgf3.1 vector to generate plasmid Xfgf3.9. Thus,
COS-1 cells transfected with Xfgf3.7 or Xfgf3.8 contained gp31, plus a
trace of the smaller form that presumably reflects cleavage at R235. In
the absence of heparin, there was no evidence for an
NH2-terminally processed form of XFGF3 in either the cell
extracts or the medium (Fig. 6a)
. The medium from
Xfgf3.7-transfected cells was therefore used to test the biological
activity of unprocessed XFGF3. However, in presence of heparin,
NH2-terminally processed products were detected in the
conditioned medium from cells expressing both Xfgf3.7 and Xfgf3.8 (Fig. 6b)
. Because these forms were not observed in the cell
extracts, they presumably arose by extracellular cleavage of the
already secreted proteins and may be attributable to conformational
changes induced by heparin, favoring the cleavage by proteases at this
specific site. The mutations would not have prevented cleavage by
trypsin-like activities in the medium.
|
, even when used undiluted (not shown). The same was true for the Xfgf3.8
conditioned medium (Fig. 7a)
. In contrast, the deletion
mutant where most of the part between the signal peptide cleavage site
and the RQRR motif were deleted was highly transforming on NIH3T3 cells
(Fig. 7a)
.
|
, but there was a marked difference in their activity on
C57MG cells (Fig. 7c)
. Although the medium from
Xfgf3.1-transfected COS-1 cells caused a dose-dependent increase in
[3H]thymidine incorporation in quiescent C57MG cells
(Fig. 7c)
, the Xfgf3.7-conditioned medium achieved <10% of
the maximal incorporation observed with the wild-type protein. However,
medium from COS-1 cells transfected with the deletion mutant Xfgf3.9
were highly mitogenic on both cell lines. Taken together, these data
imply that the unprocessed form of XFGF3 might be impaired in its
ability to interact with the receptor isoforms expressed on C57MG
cells.
Receptor Binding Analysis on COS-1 Cells Expressing
Xenopus FGF Receptors.
To further explore receptor binding, we used the Xfgf3.1, Xfgf3.7, and
Xfgf3.9 conditioned media to purify the respective cleaved and
uncleaved forms of XFGF3 on heparin-Sepharose. The proteins were eluted
using a discontinuous NaCl gradient, and fractions containing the
XFGF3-related proteins were identified by immunoblotting. The relative
amounts of XFGF3 in the recovered fractions were determined by
immunoblotting of serially diluted samples, and equivalent amounts were
then tested for their ability to bind to different Xenopus
FGFR isoforms. COS-1 cells were transfected with pKC3-based expression
vectors encoding XFGFR1 (IIIc), XFGFR2 (IIIbß), and
XFGFR2(IIIcß). Because FGF1 binds with high affinity to all known
FGFRs a competition assay was established in which the different
ligands were used to compete with 125I-labeled FGF1 for
binding to each receptor isoform (32
, 33
, 43)
. Thus, COS-1
cells expressing the different FGFRs and the vector-only control were
incubated with 125I-labeled FGF1 in the presence or absence
of the different forms of purified XFGF3 or with unlabeled FGF1 as
competitor. After covalent cross-linking, the labeled receptor-ligand
complexes were fractionated by SDS-PAGE and detected by autoradiography
(Fig. 8a)
. In the absence of competitor, appropriate bands were detected
corresponding to the two immunoglobulin loop, IIIb and IIIc isoforms of
XFGFR2 at Mr 170,000 and to the three
immunoglobulin loop form of XFGFR1 at Mr
150,000. In the presence of unlabeled FGF1, there was no detectable
binding of 125I-labeled FGF1 to any of the receptors.
Although mature XFGF3, purified from Xfgf3.1-transfected cells,
competed for binding to all three receptor isoforms, the uncleaved form
of XFGF3 recovered from Xfgf3.7-transfected cells was clearly less
effective in this assay.
|
.
|
. The higher molecular weight species of the glycosylated forms and
the nonglycosylated protein were shown previously to still retain the
signal peptide, suggesting that the signal peptide cleavage of FGF3 is
less efficient. Similar different intracellular forms resulting from
glycosylated FGF3 forms before and after signal peptide cleavage can be
detected with plasmids containing the wild-type cDNA encoding the CUG
initiated NH2-terminally extended form. However, only a
single form of FGF3 (gp32.5), which is endopeptidase H resistant, is
secreted. Because in their normal context the signal peptide cleavage
site and the RLRR motif are separated by only six residues and FGF3
undergoes further carbohydrate modification in the secretion pathway,
it is difficult to demonstrate cleavage at the RLRR by comparing the
molecular size of the secreted form. Therefore, to be able to confirm a
NH2-terminal trimming of the mouse FGF3, we used a mutant
cDNA with a spacer of 15 amino acids of random structure between the
two cleavage signals. The insertion had the effect that all of the FGF3
protein is targeted into the secretion pathway (23)
. COS-1
cells were transfected with pKC3.2 and pKC4.12
15. Cell extracts and
supernatant of the transfected cells were analyzed by SDS-PAGE and
immunoblotting using a specific COOH-terminal antipeptide serum against
mouse FGF3. No FGF3-related proteins were detected in cells transfected
with the vector alone, whereas the pKC3.2 vector generated the expected
major FGF3 products. With the vector containing the mutant cDNA, the
predominant intracellular products were about Mr
32,000–33,000, as expected for the nonglycosylated and noncleaved form
and the glycosylated form after signal peptide cleavage. Small amounts
of Mr 35,000 and Mr
30,000 species were also observed, which we interpret as the noncleaved
glycosylated form and the cleaved but not gylcosylated form of
FGF315. Analysis of the medium revealed that with both plasmids,
only a product of the same size was present in the supernatant. Because
the antibody was directed against the COOH-terminal 15 amino acids,
this result suggests that the secreted product of FGF315 must have
undergone additional processing at its NH2-terminus. To
show that the protein backbone of both secreted FGF3 forms have the
same size, FGF3 products were therefore recovered from pKC3.2- and
pKC41215-transfected COS-1 cell conditioned medium by
immunoprecipitation with the mouse FGF3-specific polyclonal serum and
treated with neurominidase and N-glycanase to remove the
carbohydrate. The products were then fractionated by SDS-PAGE and
detected by immunoblotting. Digestion with the glycosidases reduced the
size of the secreted products to exactly the same size, confirming the
NH2-terminal trimming of 4.1215 (Fig. 9)
. Discussion
We previously described the isolation and characterization of
cDNAs encoding the Xenopus homologue of FGF3, but there were
several unresolved issues relating the pseudotetraploid nature of
the Xenopus genome (37)
. In X.
laevis embryos, two major transcripts can be detected from the
early gastrula stage onward (44)
. These RNAs could be
generated from a single gene by differential promoter usage and
polyadenylation, as documented for the mouse Fgf-3 gene
(11
, 12
, 16)
or represent transcripts from separate genes.
Given the relatively low abundance of these RNAs in stage 17 embryos
(44)
, only four independent cDNAs were identified in the
original library screen and because none of them contained a
polyadenylation site, it was not possible to relate them to different
transcripts. However, DNA sequencing and PCR analysis of genomic
DNA clearly indicate that the cDNAs represent separate alleles (see
Fig. 1
).
In view of this, it became important to establish whether either of the
alleles was capable of encoding XFGF3 isoforms more closely resembling
the mouse counterpart. Mouse Fgf-3 encodes two primary
translation products that initiate either at a CUG codon or at a
downstream, conventional AUG codon (20
, 23)
. The sequence
of human FGF3 predicts a similar situation
(36)
. Whereas the AUG-initiated form of mouse FGF3 is
exclusively directed into the secretory pathway, the fate of the
CUG-initiated form is determined by competition between signals for
nuclear localization and secretion (23
, 34)
.
Overexpression of the nuclear form inhibits cell proliferation and
leads to a G1-phase cell cycle arrest (45)
,
whereas secreted FGF3 is mitogenic for cells expressing the appropriate
receptors (21
, 33)
. Significantly, none of the
Xenopus cDNAs characterized here encoded an
NH2-terminally extended form of the protein because the
presumed AUG initiation codon is preceded by multiple in-frame
termination codons (Fig. 1a)
. A similar situation prevails
in Zebrafish FGF3, suggesting that the nuclear form of mouse FGF3
evolved after the divergence of mammals and lower vertebrates
(43)
. However, the Zebrafish gene and both of the
Xenopus alleles specify an additional segment of 18 amino
acids that are not present in mouse and human FGF3.
Because the Xenopus homologue of FGF3 is more widely expressed and can interact with a broader range of receptors than its mammalian counterparts (15 , 25 , 32 , 44 , 46) , it is interesting to consider whether mechanisms for generating biological diversity from a single ligand, such as endoproteolytic cleavage, have been supplanted during evolution by the creation of separate genes with more specialized roles. In this study, we provide unequivocal evidence that XFGF3 is subject to NH2- and COOH-terminal cleavage that is reminiscent of the processing of active proteins from their inactive or less active precursors. Thus, the cleavage occurs after arginine residues that conform to the general consensus (R/K)-Xn-(R/K). Such sites can be found in the precursors of a variety of proteins, including polypeptide hormones (e.g., prorenin, proinsulin, and pro-parathyroid hormone), growth factors (e.g., BMP-4, pro-TGFß, pro-NGF, and pro-neurotensin 3), and growth factor receptors (e.g., insulin receptor; Refs. 40 and 47 ). Using a FLAG-tagged mutant of XFGF3, where the recognition epitope is directly after the presumed RQRR cleavage motif, we demonstrated the existence of intracellular forms of XFGF3 that were detectable with the M1 monoclonal antibody, which only recognizes the FLAG epitope as a free NH2 terminus.
By analogy to other processed proteins, intracellular cleavage of XFGF3
is likely to occur in the Golgi, and the FLAG-tagged version of gp27 is
clearly detectable in this compartment (Fig. 5)
. In mammalian cells, a
number of serine proteases have been identified, termed PCs, that are
related to the yeast protein kexin and cleave precursors adjacent to
pairs of basic residues (48
, 49)
. Whereas PC1 and PC2
recognize simple pairs of basic residues, others, such as furin and PC7
which reside in the trans-Golgi network of most cell types,
prefer to cleave after the general motif (R)-X-(R/K)-R (40
, 50)
. These or related enzymes are therefore strong candidates
for the intracellular processing of XFGF3, which occurs adjacent to
NH2- and COOH-terminal RQRR motifs.
Despite clear evidence for intracellular processing of XFGF3, a
considerable amount of the uncleaved gp31 protein is exported into the
extracellular space (Figs. 2
and 5)
. Although this form can persist in
the ECM, displacement into the medium after addition of soluble heparin
results in efficient processing to gp27. Using plasminogen as a source
of active plasmin (51)
, we were able to formally
demonstrate the precursor-product relationship of gp31 and gp27.
Plasmin exhibits a broad trypsin-like substrate specifity and degrades
several ECM components such as fibronectin, laminin, and the protein
core of proteoglycans (52)
. It has also been shown to play
a major role in the mobilization of FGFs from ECM, and cleavage of the
gp31 isoform of XFGF3 may be part of this release process
(53)
. Alternatively, because plasmin activates secreted
latent TGF-ß1, we considered the possibility that the conversion of
the secreted gp31 to gp27 might potentiate its biological activity.
By altering the RQRR motif, we produced mutant forms of XFGF3 that did not undergo intracellular cleavage, and by growing cells in the absence of heparin, were able to obtain conditioned medium containing the uncleaved protein. This form of XFGF3 proved significantly less active than mature gp27 in its ability to induce morphological transformation of NIH3T3 cells and in its mitogenic activity on quiescent C57MG cells but did not appear to be functionally compromised when tested on HC11 cells. This presumably reflects the different receptor isoforms expressed by different cells types because C57MG cells express predominantly the FGFR1 IIIc receptor isoform, whereas HC11 express mainly the IIIb isoform of FGFR2 (31 , 33 , 54 , 55) . Recent studies have shown that although mouse, Zebrafish, and Xenopus FGF3 show the same high affinity for FGFR2(IIIb), Xenopus FGF3 also binds with high affinity to FGFR2(IIIc) and with substantially lower affinity to the FGFR1(IIIc) isoform (32) . By comparison to the wild-type protein, the unprocessed form of XFGF3 encoded by Xfgf3.7 had a reduced affinity for all Xenopus FGFR variants tested.
Therefore, the lower mitogenic activity on C57MG cells and lack of transforming activity on NIH3T3 cells could be explained by a threshold effect rather than a distinct receptor specificity for the unprocessed form. Interestingly, proteolytic processing of the TGF-ß precursor at a similar motif R-H-R-R as those found in the Xenopus FGF3 protein resulted in a 5-fold increase in its biological activity (56) . Furthermore, blocking the proteolytical activation of BMP-4 a member of the TGF-ß family during early Xenopus embryonic development phenocopies the lost of BMP-4 activity (47) . It was reported recently that NH2-terminal truncation of FGF4, which removes the single N-linked glycosylation site, caused a marked increase in receptor binding activity (57) . The authors argued that the NH2-terminal region of FGF4 may have a flexible and disordered structure, as suggested by the analysis of the three-dimensional structure of FGF1 and FGF2, and that removal of the NH2-terminus results in a more organized form contributing to the stability of the ligand-receptor interaction (57) . Significantly, the NH2-terminal trimming of Xenopus and mouse FGF3 removes in each case a putative turn-coiled structure, supporting this interpretation. Therefore, regulation of precursors by proteolytic processing appears to be not a matter of switching from an inactive form to an active form rather than from a less active to a more active growth factor isoform. During evolution, the reason for the NH2-terminal processing may have changed. Whereas in Xenopus isoforms of FGF3 with different affinities or stability may be necessary because of a restricted set of different FGFs, in mammals FGF3 had gained a new function as a nuclear signalling factor via a sophisticated NH2-terminal architecture of nuclear and secretion pathway targeting motifs that made necessary an NH2-terminal trimming of the secreted form.
Taken together, as we have demonstrated previously for mouse, Zebrafish, and Xenopus FGF3 that their COOHterminal domain governs the receptor specificity (24 , 25) , here, we provide the first evidence that the NH2-terminal processing appears to be essential for high receptor binding affinity.
Materials and Methods
Isolation and Characterization of Xfgf3 cDNA Clones.
A stage 17 X. laevis embryo library, originally constructed
in the laboratory of D. Melton, was screened by hybridization with a
290-bp AluI fragment that encompasses the second exon of
mouse Fgf-3 genomic DNA (58)
. A total of 3 x 105 phage plaques, immobilized on Hybond N filters
(Amersham), were hybridized at 37°C for 16 h in 4 x SSC
(1x SSC is 0.15 M NaCl and 0.015 M sodium
citrate), 40% formamide, 0.04% each of bovine serum albumin, Ficoll
and polyvinylpyrrolidone, and 50 µg/ml yeast tRNA. The filters were
washed at 50°C in 2x SSC and 0.1% SDS, and the signals were
detected by autoradiography. Phage DNA was recovered from liquid
lysates by standard procedures, and individual EcoRI
fragments were transferred into plasmid vectors for further analysis
and DNA sequencing. DNA sequencing was accomplished by
chain-termination procedures using double-stranded DNA templates and
Sequenase protocols as recommended by the supplier.
PCR Amplification of Genomic DNA.
Samples (1 µg) of high molecular weight DNA from the blood of a
single adult X. laevis were subjected to amplification by
PCR using primers derived from the Xfgf3 cDNA sequence. The
5' primers were based on nucleotides 38–64 and 286–306 in the
sequence shown in Fig. 1
. The 3' primer corresponded to a highly
conserved region in the first exon (nucleotides 487–508).
Amplification was performed as follows: 95°C for 2 min, 55°C for 2
min, and 72°C for 3 min. The products were analyzed by
electrophoresis in a 1% agarose gel and visualized by staining with
ethidium bromide.
Cell Culture and DNA Transfection.
COS-1 cells were maintained in DMEM containing 10% FCS and passaged
every 7 days at a ratio of 1:10. For DNA transfection, 20 µg of
CsCl purified plasmid DNA were introduced into 5 x
105 COS-1 cells by electroporation using a Bio-Rad
Gene-Pulser set at 450V/250 µF. Cells were grown in 90-mm Petri
dishes, and between 48 and 72 h after transfection, they were
harvested for protein analysis. In some experiments, heparin was added
12 h after transfection at a concentration of 5 µg/ml of the
culture medium.
Construction of Plasmids.
Vectors expressing mouse FGF3 (pKC3.2) and Xenopus
FGF3 (Xfgf3.1) in an SV40-based expression vector (pKC) have been
described previously (19
, 25)
. Synthetic oligonucleotides
were used as primers for amplification and modification of the plasmid
Xfgf3.1 by PCR. As a first step, a new SacII site was
created by PCR, which changed codons 41 and 42 (from CCCAGG to CCGCGG)
but not the encoded amino acids. For plasmid Xfgf3.7, a second PCR
product was amplified that had this SacII site as well as an
altered codon 45 (CGA to AAA) to replace R with K. In Xfgf3.6, the 5'
primer
(5'-CCGCGGCAGAGACGAGATTATAAGGATGACGATGACAAAGATGCTGGGGGACGTGG-3')
was designed to introduce the FLAG epitope DYKDDDDK adjacent to the
SacII site (underlined). The 3' primer
(5'-CTGCAGTCACAGGTCCTCCTCGCTGATCAGCTTCTGCTCCTGTCCTCGTCTTTGTCTTCCGCTCGGTTT-3')
was designed to include 10 codons encoding an epitope from human c-MYC
recognized by the monoclonal antibody 9E10 and a unique PstI
site immediately adjacent to the translational stop codon. The PCR
fragment was fused to the 5' fragment of Xfgf3.1 using the newly
created SacII site. The reassembled cDNA was then ligated
between the XbaI and PstI sites in the SV40-based
expression vector, pKC3. For Xfgf3.8, primers were designed to
separately amplify the 5' and 3' sequences of Xfgf3.1 and create a new,
unique XhoI site at codons 45 and 46. The reverse primer for the NH2-terminal segment was CTCGAGTCCCTGCCTGGGGTCGCA,
and the forward primer for the carboxy-terminal segment was
CTCGAGGCTGGGGGACGTGGA. The PCR products were ligated through the new
XhoI site, resulting in a changed amino acid sequence from
RQRRD to RQGLE. In Xfgf3.9, the Leu24 codon in Xfgf3.1 changed from CTG
to CTC to create a unique XhoI site, and a 3' fragment of
Xfgf3.1 was amplified using the 5' primer CTCGAGCCCAGGCAGAGACGAGAT and
fused to the 5' fragment through the new XhoI site to delete
codon 26 up to codon 40.
Purification of Xenopus FGF3-related Proteins.
COS-1 cells were transfected with Xfgf3.1 or Xfgf3.7, and after 12 h, the medium was replaced with DMEM containing 0.1% FCS.
Approximately 48 h later, the conditioned medium was recovered,
and phenylmethylsulfonyl fluoride was added to a final concentration of
1 mM. 0.5 ml of a 50% slurry of heparin-Sepharose beads
(Pharmacia) was added to the medium and mixed slowly overnight at
4°C. The beads were then poured into a column (Bio-Rad PolyPrep) and
washed with 10 ml of PBS. The bound protein was eluted with a step
gradient from 0.5 to 2.0 M NaCl in PBS. Fractions were
collected, and the XFGF3-related protein-containing fractions were
determined by immunoblotting. The amount of the mutant XFGF3 proteins
was estimated against a defined amount of XFGF3 by serial dilution and
by immunoblotting using the MSDI monoclonal antibody, which recognizes
all naturally occurring XFGF3 variants and all recombinant XFGF3
proteins.
Thymidine Incorporation Assay.
C57MG mammary epithelial cells were grown routinely in DMEM containing
10% newborn calf serum. HC11 cells were grown in RPMI 1640
supplemented with 10% FCS, 5 µg/ml insulin, and 10 ng/ml epidermal
growth factor (59)
. For mitogenicity assays, cells were
transferred to 48-well tissue culture plates in 500 µl of medium.
After 24 h, the medium was replaced by DMEM containing 0.1%
serum, and after an additional 72 h, the cells were treated with
the test samples in fresh DMEM containing 0.1% serum. DNA synthesis
was assayed 12 h later by labeling the cells with 2 µCi of
[3H]thymidine/well for a further 4 h and measuring
the incorporation of label into acid-insoluble material. Cells were
washed three times in PBS, twice in methanol for 5 min, and dehydrated
in water before fixing in cold 5% trichloroacetic acid for 20 min. The
cells were finally dissolved in 300 µl of 0.3 M NaOH, and
the residual 3H-labeled material was quantified by liquid
scintillation counting.
Immunoblotting and Immunofluorescence.
The procedures used for preparing cell lysates, ECM, and conditioned
medium have been described in detail elsewhere (21)
.
Samples from equivalent numbers of cells were fractionated by SDS-PAGE
in 12.5% gels, transferred to nitrocellulose membranes (Schleicher and
Schuell), and then processed with the monoclonal antibody (9E10), which
recognizes the MYC tag, the anti-FLAG monoclonal antibodies M1 and M2,
the polyclonal COOH-terminal antibody X3, or the monoclonal antibody
(MSDI), which recognizes an epitope in Xenopus FGF3. Immune
complexes were detected with 125I-labeled protein-A
(Amersham) as described (21)
or by ECL as described by the
manufacturer. For immunofluorescence microscopy, transfected COS-1
cells were grown on glass coverslips and fixed in 4% paraformaldehyde
in PBS for 20 min. The cells were then permeabilized with 0.2% Triton
X-100 for 4 min and treated with 3% BSA in PBS (BSA-PBS). The
coverslips were then exposed to primary antibodies, and fluorescently
labeled secondary antibodies were diluted in BSA-PBS. Confocal images
were taken on a Zeiss confocal laser scan microscope.
Iodination.
FGF1 (TEBU) was iodinated by the chloramine-T method as described
(32)
. The labeled proteins were purified by heparin-sepharose
chromatography as described previously (43)
.
FGFR Competition Binding Assay.
Vectors for the expression of Xenopus FGFR1(IIIC),
FGFR2(IIIb), and FGFR2(IIIc) cDNAs in COS-1 cells have been described
previously (32)
. COS-1 cells were transfected with the
appropriate Xenopus FGFRs and seeded at 5 x
104 cells/well into 48-well plates pretreated with
poly-L-lysine (Sigma) as recommended by the manufacturer.
After 48 h, the cells were washed twice with the ice-cold binding
medium (DMEM containing 50 mM HEPES, pH 7.4, and 1 mg/ml
BSA) and incubated for 3 h at 4°C with the indicated amounts of
125I-labeled FGF1 in binding medium. Competition binding
was performed in the presence of excess of unlabeled ligands. The cells
were then rinsed twice with cold binding medium, solubilized in 0.1%
SDS, 0.3 M NaOH for 30 min at 37°C, and the associated
gamma radiation was counted. To determine the specific binding, the
radioactivity bound to cells transfected with empty vector was
subtracted from that of cells receiving FGFRs.
Cross-Linking of Receptors and FGFs.
COS-1 cells were transfected with different Xenopus FGFRs
and were cultured for 48 h in 60-mm culture dishes. The assay was
essentially performed as described previously (32
, 43)
.
Briefly, the cell monolayers were washed twice with ice-cold binding
medium and incubated for 2 h at 4°C with binding medium
containing iodinated FGF1 in the presence or absence of unlabeled
ligand or vector alone. After washing, the cells were treated for 20
min at 4° with 0.3 mM disuccinimidyl suberate (Pierce) in
PBS. After cross-linking, the cells were washed and lyzed. The extracts
were then analyzed by SDS-PAGE on 7.5% gels, and the receptor-ligand
complexes were detected by autoradiography.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 To whom requests for reprints should be
addressed, at Heinrich-Heine-Universität Düsseldorf
Medizinische Fakultät, IHTM Moorenstrasse 5, D-40225
Düsseldorf, Germany. Phone: 49-211-811-8793; Fax.:
49-211-811-6649; E-mail: kiefer{at}med.uni-duesseldorf.de 
2 The abbreviations used are: FGF, fibroblast
growth factor; ECM, extracellular matrix; FGFR, FGF receptor;
ID50, 50% displacement; PC, proprotein convertase; TGF,
transforming growth factor.
Received for publication 6/ 7/00. Revision received 9/19/00. Accepted for publication 9/25/00.
References
This article has been cited by other articles:
 |
|
 |
|
  Y. Kliger, E. Gofer, A. Wool, A. Toporik, A. Apatoff, and M. Olshansky Predicting proteolytic sites in extracellular proteins: only halfway there Bioinformatics, April 15, 2008; 24(8): 1049 - 1055. [Abstract] [Full Text] [PDF]
|
|
| ||||
