New Vascular Endothelial Growth Factor Isoform Generated by Internal Ribosome Entry Site-Driven CUG Translation Initiation (original) (raw)

Abstract

We recently demonstrated that the very long 5′-untranslated region (5′-UTR) of the vascular endothelial growth factor (VEGF) mRNA contains two independent internal ribosome entry sites (IRES A and B). In the human sequence, four potential CUG translation initiation codons are located in between these IRES and are in frame with the classical AUG start codon. By in vitro translation and COS-7 cell transfections, we demonstrate that a high mol wt VEGF isoform [called large VEGF (L-VEGF)] is generated by an alternative translation initiation process, which occurs at the first of these CUG codons. Using a bicistronic strategy, we show that the upstream IRES B controls the translation initiation of L-VEGF. This isoform is 206 amino acids longer than the classical AUG-initiated form. With a specific antibody raised against this NH2 extension, we show that the L-VEGF is present in different mouse tissues or in transfected COS-7 cells. We also demonstrate that L-VEGF is cleaved into two fragments: a 23-kDa NH2-specific fragment and a fragment with an apparent size similar to that of the classical AUG-initiated form. This cleavage requires the integrity of a hydrophobic sequence located in the central part of the L-VEGF molecule. This sequence actually plays the role of signal peptide in the classical AUG-initiated form. The AUG-initiated form and the COOH cleavage product of the L-VEGF are both secreted. In contrast, the large isoform and its NH2 fragment present an intracellular localization.

These data unravel a further level of complexity in the regulation of VEGF expression.

VASCULAR ENDOTHELIAL GROWTH factor (VEGF) is a potent mitogen of endothelial cells and plays a crucial role in the regulation of both physiological and pathological angiogenesis. VEGF is involved not only in embryonic development and differentiation of the vascular system, wound healing, and reproductive function but also in pathological angiogenic processes such as proliferative retinopathies, tumor growth, arthritis, and psoriasis (for review see Ref. 1).

Numerous studies have been devoted to understanding the expression regulation of this factor, especially at the transcriptional level. A wide range of cytokines or oncogenic proteins including IL-1β (2), IL-6 (3), IGF-I (4), TGFβ (5), c-Src (6), v-Raf (7), or Ras (8), together with oxygen pressure (9), have been shown to regulate VEGF gene transcription. VEGF expression is also posttranscriptionally regulated. Human VEGF pre-mRNA undergoes alternative splicing, which generates five polypeptide isoforms of 121, 145, 165, 189, and 206 amino acids (a.a.), the functions of which have not yet been fully defined (10, 11). VEGF mRNA stability is also influenced by hypoxic conditions (12). Posttranslational modifications of VEGF isoforms, including plasmin proteolysis or glycosylation, have been described (13–15). Recently, we and others have demonstrated a translational regulation of VEGF expression (16–19). The VEGF mRNA, with the c-myc mRNA, are the first messengers described so far which bear two independent internal ribosome entry sites (IRES A and B) in their 5′-untranslated region (5′-UTR) (16). The VEGF IRES A located upstream from the AUG codon directs a cap-independent translation initiation, which has been shown to allow VEGF synthesis in hypoxic conditions (17). VEGF thus belongs to the growing family of growth factor-coding messengers, the translation of which is controlled by an IRES-dependent mechanism (16, 20–22). The expression of many of these genes is also regulated by an alternative translation initiation process that allows the synthesis of different protein isoforms from a single mRNA; this alternative initiation occurs mainly at the level of the AUG and CUG codons (23–27). This mechanism is of major importance as it generates isoforms with distinct cellular localizations and functions. In the case of the angiogenic fibroblast growth factor 2 (FGF-2), four nuclear isoforms initiated at four distinct CUG codons and a cytoplasmic AUG-initiated form are generated from the unique FGF-2 mRNA (27).

Here we provide evidence of further VEGF isoform diversity by showing that an alternative translation initiation process, IRES B driven, occurs at the first upstream CUG codon. This generates a larger VEGF isoform bearing an NH2-terminal extension of 206 a.a. Interestingly, we demonstrate that a high mol wt form of VEGF [large VEGF (L-VEGF] is cleaved into two polypeptides. The COOH product has the same apparent size as the classical AUG-initiated form and is secreted from the cell. In contrast, the NH2 product has an intracellular localization. Our data thus reveal a new level of regulation of VEGF production.

RESULTS

The human VEGF mRNA possesses a 1,038 nucleotide (nt) long 5′-UTR. This so-called “untranslated region” actually contains an unusually large open reading frame (ORF). Remarkably, four potential CUG initiation codons are located at the 5′-end of this ORF and are located in frame with the canonical AUG start codon (Fig. 1A). Three of these CUG codons are localized between the two internal ribosome entry sites (IRES) that we have recently described (16). We therefore investigated the possibility that an extended form of VEGF protein could be synthesized from one of these codons. We also address the question of the possible role of IRES B (nt 98 to 475 in the 5′-UTR) in this potential alternative translation initiation process.

Figure 1.

The 5′-UTR of VEGF mRNA Is Highly Conserved in Mammals Alignment of VEGF a.a. sequences deduced from the ORF including the ATG codon in human, bovine, and mouse cDNA, respectively. An arrow shows the starting methionine of the VEGF 165 isoform. A double arrow indicates the cleavage site of the VEGF signal peptide. Divergent a.a. are boxed in white or in gray when the global charge is conserved. The mouse sequence used in this alignment is available from GenBank.

Structural Evidence for the Existence of a L-VEGF Protein Isoform

We first compared the three available VEGF 5′-UTR nt sequences, i.e. human, bovine, and mouse. In both the bovine and human sequences, all four CUG codons were found to be conserved and located in frame with the AUG (28, 29). In contrast, only the first CUG was conserved in the mouse 5′-UTR, and this codon was not in frame with the AUG (30). Surprisingly, the nt sequences upstream from the AUG were highly conserved in all three species (not shown). We cloned and sequenced again the mouse VEGF 5′-UTR and found marked discrepancies with the published sequence. The 5′-UTR sequence that we obtained (GenBank accession no. in progress) was actually very similar to that of the human and bovine sequences with five CUG codons in frame with the AUG initiation codon.

Comparison of the deduced a.a. sequences in the three species (Fig. 1B) revealed that a 358- to 361-a.a. protein could be encoded by the mouse, bovine, and human VEGF 165 cDNAs, respectively. Moreover, the a.a. identity of this putative amino-terminal extension of VEGF was high (76%) in all three species and could be compared with that observed in the classical AUG-initiated VEGF165 protein (87%). This degree of conservation, which is unexpected from a proposed “uncoding” sequence, led us to investigate whether the region located downstream from the CUG could represent a real coding sequence and generate a larger VEGF isoform.

In the case of the rat sequence (31), no ORF is apparent. This could result from a genetic divergence, but it should be noted that this region is very GC rich and some technical problems during cDNA cloning or during sequence determination could also occur similar to those observed for the mouse (present paper). Partial 5′-VEGF sequences are also available for zebrafish, Xenopus, and quail but no ORF is evident at this time.

VEGF cDNA Generates a L-VEGF Isoform in Vitro

We first analyzed the products obtained after translation in the in vitro rabbit reticulocyte lysate system of an in vitro transcribed VEGF189 mRNA, which contained or did not contain the 5′-UTR. Translation of the full-length VEGF mRNA produced two main proteins with apparent sizes of 43 kDa and 23 kDa, respectively (Fig. 2A, lane 1). The 23-kDa product corresponded in size to that of the AUG-initiated isoform (Fig. 2A, lane 2). The larger and more abundant translation product had the expected size of a CUG1-initiated protein.

Figure 2.

Detection of the L-VEGF Protein A, In vitro translation of the VEGF 189 mRNA with or without 5′-UTR, transcribed from the 5′189 or the 189 constructs, respectively. Lanes 1 and 2 correspond to nonimmunoprecipitated (NI) samples. In lanes 3 and 4, the 5′189 translation products were immunoprecipitated with the anti-VEGF or the N-Ab antibodies, respectively. B, Western blotting using the N-Ab on extracts of COS-7 cells transfected with 5′189 (lane 1) or 189 (lane 2) or untransfected (NT, lane 3). C, Detection of the endogenous L-VEGF protein in different mouse tissue extracts by Western blotting using the N-Ab. Lane 5′189 corresponds to an extract of COS-7 cells transfected with the 5′189 construct. The following lanes correspond to mouse lung (Lu), brain (Br), liver (Li), spleen (Sp), and kidney (Ki) protein extracts, respectively. The arrow indicates the position of the L-VEGF isoform. D, Detection of the endogenous L-VEGF protein in different mouse tissue extracts by Western blotting using the VEGF-Ab. Lane 5′189 corresponds to an extract of COS-7 cells transfected with the 5′189 construct.

To confirm that the large protein actually corresponded to a VEGF isoform, a polyclonal antibody (N-Ab) was raised against a 121-a.a. peptide deduced from the 5′-UTR sequence limited by nt 553–917, located between the CUG1 and the AUG codons. As shown in Fig. 2A, lane 4, this N-Ab only immunoprecipitated the larger 35S-labeled protein whereas an anti-VEGF antibody (VEGF-Ab), used in parallel, immunoprecipitated both isoforms (lane 3). These data suggested that the large translation product protein corresponded to a VEGF isoform initiated upstream from the canonical AUG.

The production of this large isoform was then tested in COS-7 cells. These cells were transfected with the VEGF189 or the VEGF-5′189 constructs shown in Fig. 2A. Cell extracts were analyzed by Western blotting using the N-Ab. As shown in Fig. 2B, a large 43-kDa isoform was specifically detected in cells transfected with the cDNA containing the 5′-leader. It should be noted that N-Ab specifically detects VEGF products initiated upstream from the AUG codon but not an AUG-initiated protein (Fig. 2B, lane 189). We also detected a major polypeptide around 23 kDa in 5′189-transfected cells (see further explanation in Fig. 5A).

Figure 5.

Processing of the L-VEGF Isoform A, Upper panel, Schematic representation of the constructs transfected in the COS-7 cells. In construct 1, chimeric VEGF-CAT gene was fused to the 5′-UTR. Construct 2 corresponds to the complete VEGF 189 cDNA. In construct 3, only nt 499–1,038 of the 5′-UTR were present. In construct 4, nt 499–1,038 of the 5′-UTR were fused with the VEGF signal peptide coding sequence. Lower panel, Immunodetection using the N-Ab of the proteins obtained after transfection of constructs 1–4 in COS-7 cells. B, Mutation of the initiation codon ATG1039 into noninitiation codon in the chimeric VEGF-CAT gene. The VEGF 5′-UTR is absent from constructs 1 and 2. In constructs 5 and 6, the CTG499 was mutated into a strong initiation codon ATG. Proteins were detected by Western blotting using the CAT-Ab after transfection of COS-7 cells with constructs 1–6. The numbers of the constructs used are reported below the lanes.

Detection of the L-VEGF Form in Vivo

To verify the occurrence of the L-VEGF form _in viv_o, we performed Western blotting on mouse brain, liver, spleen, kidney, and lung tissue extracts using the N-Ab (Fig. 2C) or the VEGF-Ab (Fig. 2D).

The presence of a large 43-kDa immunoreactive protein was observed in all the tissue extracts with both antibodies with the same relative intensities. This large protein has the same electrophoretic mobility as that of the L-VEGF protein produced in COS-7 cells transfected with the full-length VEGF189 cDNA (5′189 in Fig. 2, C and D). In Fig. 2D, the 21- and 23-kDa proteins correspond to VEGF189 AUG-initiated forms.

This experiment suggests the physiological reality of the endogenous synthesis of a L-VEGF isoform.

Identification of the L-VEGF Translation Initiation Codon

To further characterize the translation initiation codon, we studied the effect of different deletions or point mutations in the 5′-UTR on L-VEGF translation. These experiments were performed both in vitro and in transfected COS-7 cells.

We chose to use a reporter gene that encodes a chimeric VEGF-chloramphenicol acetyltransferase (CAT) protein. This protein is indeed easily detectable with an anti-CAT antibody. The use of this reporter also avoided any confusion due to the presence of endogenous VEGF in the transfected cells. The reporter cDNA consisted of fusion of the CAT reporter gene 168 nt downstream from the VEGF AUG. The different chimeric VEGF-CAT constructs used are shown in Fig. 3A.

Figure 3.

Identification of the L-VEGF Translation Initiation Codon A, Schematic representation of the VEGF-CAT constructs containing different parts of the 5′-UTR region. The first nt of the construct is identified by its position in the full 5′-UTR (see Materials and Methods). In constructs D and E, the first CTG codon was mutated into a strong ATG initiation codon or a noninitiating TTT codon, respectively. B, Analysis of proteins translated in vitro from the messengers transcribed from the constructs shown in Fig. 3A. The first lane corresponds to the reaction in the absence of mRNA. C, Immunodetection using an anti-CAT antibody (CAT-Ab) of the translation products obtained after transfection of the constructs shown in Fig. 3A in COS-7 cells. The names of the constructs transfected are indicated below the lanes. The first lane corresponds to the untransfected cells extract. The last two lanes show an immunodetection with the CAT-Ab on extracts of COS-7 cells transfected with construct A, treated or not with tunicamycin.

Figure 3B shows the translation products obtained in a reticulocyte lysate with the messengers transcribed from these constructs. Lane C corresponded to the translation of the messenger containing the full-length 5′-UTR (construct C). The L-VEGF-CAT protein was clearly produced but the low molecular mass AUG-initiated form (32 kDa) was barely detectable. The 50-kDa L-VEGF-CAT protein was not detectable when the mRNA lacked the CUG1 (Fig. 3B, lanes A and B).

This confirms the role played by the first CUG in such synthesis. This is further strengthened by the results obtained with construct E. Indeed, the mutation of CUG1 into a noninitiating UUU codon completely abolished translation of the 50-kDa L-VEGF-CAT (lane E). The additional isoform observed in that case could correspond to a CUG2-initiated protein. This has already been observed for other alternative initiation processes (27, 32).

In contrast, the mutation of CUG1 into an AUG initiation codon resulted in an increased L-VEGF translation (construct D). Altogether, these results suggested that the translation of the L-VEGF isoform is initiated at the CUG1 codon.

Similar conclusions were derived from transfection experiments. The different cDNA (constructs A–E) were transfected in COS-7 cells (Fig. 3C). The cellular extracts were analyzed by Western blotting using an anti-CAT antibody (CAT-Ab). The mRNA lacking the 5′-leader (construct A) was translated into two polypeptides of 34 kDa and 29 kDa (lane A, Fig. 3C). These corresponded to glycosylated and unglycosylated forms, respectively, of the same protein. Indeed, only the unglycosylated 29-kDa protein could be detected after tunicamycin treatment of the transfected cells with construct A (Fig. 3C, right panel). The VEGF-CAT fusion did not contain the VEGF glycosylation site but we identified a glycosylation site in the CAT protein.

Again the synthesis of a large L-VEGF-CAT protein was detected from cells transfected with the full-length cDNA (lane C). However, in contrast with the in vitro results of Fig. 3B, this large protein accumulated at a lower level than the low molecular AUG-initiated forms.

The same expression profile was also observed after transfection of the nonchimeric VEGF189 cDNA using an anti-VEGF antibody (see Fig. 2D, 5 ′189).

Interestingly, the mutation of the CUG1 codon into an AUG codon (lane D) strongly enhanced the expression of the L-VEGF isoform. Conversely, the mutation into a noninitiating UUU codon (lane E) or the removal of the CUG1 (lane B) abolished this expression.

All together, these results confirmed that the L-VEGF protein is initiated at the CUG1 codon at position 499 in the 5′-leader.

The IRES B Directs Cap-Independent L-VEGF Translation Initiation

As the CUG1 codon is located just downstream from the IRESB in the 5′-leader, we investigated the involvement of this IRES in the L-VEGF translation initiation. To test this hypothesis, we checked whether L-VEGF initiation was cap independent and driven by IRES B in bicistronic vectors (16, 21). Two vectors, differing in the presence or absence of a stable hairpin structure upstream from the first CAT cistron, were transfected in COS-7 cells and the translation products analyzed by Western blotting using the CAT-Ab. As shown in Fig. 4, translation of the chimeric L-VEGF remained constant independently of the presence or absence of the hairpin structure that abolished cap-dependent expression of the first cistron (compare lanes 1 and 2). This suggested that IRES B located upstream from CUG1 directed the cap-independent translation initiation of the L-VEGF protein.

Figure 4.

Cap-Independent Translation of the L-VEGF Isoform in COS-7 Cells Upper panel, Bicistronic constructs transfected in the COS-7 cells. A stable hairpin structure was introduced upstream from the first CAT cistron in construct 2. Lower panel, Western blotting using the CAT-Ab on transfected COS-7 cell extracts. The positions of the CUG499-initiated VEGF-CAT protein (L-VEGF-CAT), the AUG1039-initiated VEGF-CAT isoform (VEGF-CAT), and the CAT protein encoded by the first cistron are indicated by arrows.

The L-VEGF Isoform Is Processed

As seen in Fig. 3, the ratio of L-VEGF vs. low molecular proteins that accumulated upon translation of a chimeric VEGF-CAT mRNA was markedly higher in vitro than in transfected cells.

Even more striking was the observation that the N-Ab selectively raised against the NH2 part of L-VEGF was able to recognize not only the L-VEGF but also a 23-kDa low molecular mass protein (Fig. 2B, lane 5′189) in the transfected cells.

One interpretation of these observations could be that the L-VEGF was actually cleaved physiologically in the cell and that such a cleavage did not occur in vitro at a significant level.

To test this hypothesis, chimeric VEGF-CAT and the VEGF189 cDNA were transfected in COS-7 cells, and the translation products were detected by Western blotting using the N-Ab. Transfection of both constructs led to the detection of large proteins (Fig. 5A, lanes 1 and 2). As expected, the size of these proteins varied according to the size of the coding sequences. Interestingly, two additional 23- and 24-kDa proteins were also detected regardless of the nature and the size of the coding sequence used. This supports the idea that the L-VEGF isoform is processed. The two proteins could indeed correspond to the NH2 cleavage products.

To more precisely define the cleavage site, we transfected constructs starting at the codon 499 and ending just before (nt 1,038) and just after (nt 1,115) the coding sequence for the signal peptide of the classical AUG-initiated VEGF isoform (constructs 3 and 4). In this experiment the CUG1 was mutated into an AUG to enhance the detection. The same result was obtained with the original CUG (not shown). We observed that constructs 1, 2, and 4 generated the same 23- and 24-kDa products. In contrast, construct 3, which lacked the hydrophobic peptide sequence, produced only one protein of lower molecular mass (lane 3). These results indicated that the 23- and 24-kDa polypeptides corresponded in size to proteins that contained the hydrophobic sequence corresponding to the VEGF signal peptide. We hypothesize that the 24-kDa polypeptide might correspond to an as-yet-unidentified posttranslational modification of the 23-kDa peptide since both isoforms were produced from construct 4.

The Processing of the L-VEGF Isoform Occurs at the Level of a Central Hydrophobic Sequence

The data presented above suggested that the position of the cleavage site could be at or near the end of the VEGF signal peptide. The cleavage of L-VEGF could thus generate COOH polypeptides that could migrate with the same electrophoretic mobility as the classical AUG-initiated protein. To allow the identification of this carboxy-terminal cleavage product of L-VEGF, we suppressed the synthesis of the AUG-initiated form. Three AUG codons are present in the VEGF coding sequence but only the first one, the AUG 1,039, is normally used for translation initiation. It was thus mutated into a noninitiation codon, CCC. As a first step we tested the effect of these mutations on translation of the classical low molecular mass VEGF isoform.

Figure 5B, lane 2, shows that, as expected, the mutation of the AUG completely abolished the translation initiation of these low molecular forms (compare lanes 1 and 2). We then mutated the AUG 1,039 in the full-length VEGF-CAT construct (construct 4). Despite this mutation, the accumulation of the low molecular mass proteins of 29 and 34 kDa was still observed (lane 4).

The protein of approximately 26 kDa, which appears in lane 4, most probably corresponds to a polypeptide whose translation was initiated at a downstream AUG in the VEGF sequence. When the upstream CUG codon was converted into an AUG, the L-VEGF expression was significantly increased (lanes 5 and 6). The low molecular mass proteins also clearly accumulated. However, the abundance of these proteins was not significantly different when the AUG 1,039 was converted into noninitiation codon. This was unlike what was observed in lanes 3 and 4 with the original CUG.

This suggests that these low molecular mass proteins corresponded partially (lane 3) or totally (lane 5) to cleavage products of the L-VEGF. This difference could be due to a polar effect of the upstream AUG 499, which impairs the initiation at the downstream AUG 1,039 (in lanes 5 and 6).

We thus concluded from these results that the low molecular mass proteins corresponded to the COOH fragments of the L-VEGF-CAT form. Remarkably, these two polypeptides have the same apparent molecular mass (29 and 34 kDa) as the AUG 1,039-initiated protein.

Cleavage of L-VEGF Requires a Functional Central Hydrophobic Sequence

As the cleavage of the L-VEGF occurred in the vicinity of the hydrophobic sequence that plays the role of signal peptide in the AUG-initiated form, we evaluated the influence of the functionality of this sequence in the processing of the L-VEGF isoform.

New constructs were designed in which we mutated two critical amino acids of the signal peptide (the 9th and the 14th) to impair its cleavage and the subsequent export process of the AUG-initiated form (see Materials and Methods). The effect of the mutation was evaluated by COS-7 cells transfection. Both the cellular extracts and the culture media were analyzed by Western blotting using the CAT-Ab (Fig. 6A). One notes, in Fig. 6A, lane 1, the presence of the chimeric L-VEGF as well as that of the two low molecular mass polypeptides of 29 and 34 kDa.

Figure 6.

Cleavage of L-VEGF Requires a Functional Signal Sequence A, Mutation of the VEGF peptide signal coding sequence in the chimeric VEGF-CAT cDNA. Upper panel, The mutated signal peptide is represented by a dotted box underlined (construct A′). Lower panel, Immunodetection using the CAT-Ab of the proteins obtained after transfection of constructs A and A′ in COS-7 cells. In the left panel, proteins contained in the cell extracts were analyzed (lanes 1 and 2). The right panel shows analysis of the proteins secreted in the transfected COS-7-conditioned media (lanes 3 and 4). A schematic representation of the protein detected is shown on the left of the panel. B, Mutation of the ATG1039 as well as the signal peptide (dotted box, lane 2) in constructs in which the CTG499 was replaced by a strong ATG codon. Proteins are detected by Western blotting using the N-Ab (left panel) or the CAT-Ab (right panel) after transfection in COS-7 cells.

These two last proteins were clearly exported from the cells and found in the culture medium (lane 3). The mutation of the signal peptide (construct A′) did not affect production of the L-VEGF protein in the cellular extract (lane 2) but led to the accumulation of only one 32-kDa protein. This protein was no longer exported from the cell (lane 4). The size of this protein corresponds to that of the expected unprocessed AUG 1,039-initiated form with uncleaved signal peptide (not glycosylated).

To further evaluate the effect of these mutations of the signal sequence on the processing of L-VEGF, we tested the disappearance of its amino- and carboxy-terminal cleavage products in transfected cell extracts. We introduced the mutations of the signal sequence in construct 6 of Fig. 5B in which the AUG 1,039 was converted into noninitiation codon (construct 2 in Fig. 6B). The resulting construct was transfected in COS-7 cells and analyzed by Western blotting using both the CAT-Ab and the N-Ab. As shown in Fig. 6B, lane 2 (left and right panels), in contrast with the results observed with the WT signal sequence (lane 1, left and right panels), the mutation introduced in the hydrophobic sequence abolished the accumulation of the low molecular mass products. Because this was observed with both antibodies, we concluded that the mutations that inactivated the signal peptide of the classical AUG-initiated form also prevented the L-VEGF cleavage. These results also clearly eliminated the possibility of an artifactual cleavage of the L-VEGF during cell lysis.

Cleavage of the L-VEGF Protein in Microsomes Complemented Rabbit Reticulocyte Lysate (RRL)

Because the presence of the functional hydrophobic signal sequence appeared to play a crucial role in L-VEGF processing, we tested the possibility that this processing could take place in the endoplasmic reticulum. We therefore tried to reproduce this cleavage in vitro in a reticulocyte lysate supplemented with canine pancreatic microsomal membranes.

To check the efficiency of the system, we first verified the cleavage of the signal peptide in the control _Escherichia coli_β -lactamase precursor protein (data not shown). To test the cleavage of the L-VEGF in these conditions, we used the chimeric VEGF-CAT mRNA transcribed from constructs lacking the ATG initiation codon with or without mutations in the central hydrophobic region. The neosynthesized proteins were labeled by incorporating 35S-methionine.

In uncomplemented lysate, translation of the full-length mRNA (from construct A) led to the accumulation of the 50-kDa L-VEGF protein (Fig. 7, lane A, −). In presence of microsomes, this protein was partially processed into a lower molecular mass protein (lane A,+). In contrast, as seen in lane B, the L-VEGF isoform containing a mutated hydrophobic sequence (translated from mRNA B) did not undergo such a cleavage even in the presence of microsomes. In these experiments, the NH2 fragment resulting from cleavage of the L-VEGF was not detected because it did not contain any methionine.

Figure 7.

Cleavage of the L-VEGF Protein in Microsomes Complemented RRL Upper panel, Schematic representation of the constructs used for the in vitro transcription of the mRNA used in the translation assay. The dotted box in construct B indicates the mutation of the signal peptide. In both A and B constructs, the ATG1039 was mutated into a noninitiation codon. Lower panel, Translation products labeled by incorporation of 35S-methionine. The presence or absence of microsomes in the RRL is shown by a plus or a minus, respectively, below each lane. The mRNA used in the reactions is indicated below the lanes. On the right, two arrows show the position of the L-VEGF isoform (black and white boxes) and the carboxy-terminal cleavage product (white box).

These in vitro experiments confirmed that L-VEGF processing was dependent on the presence of a functional signal peptide and occurred during the translocation into the endoplasmic reticulum.

Subcellular Localizations of L-VEGF and Its Cleavage Products

We first determined whether the L-VEGF isoform and the fragments resulting from its processing were secreted from the cells (Fig. 8A). We attempted to localize these polypeptides by transfecting COS-7 cells with full-length VEGF-CAT cDNA containing or not the mutation of the AUG 1,039. In both constructs the upstream CUG was converted into a more potent AUG initiation codon to maximize the production of the chimeric L-VEGF form and thus facilitate the localization of the products.

Figure 8.

Subcellular Localization of the L-VEGF Isoform and Its Cleavage Products A, Analysis of the secretion of the L-VEGF isoform and its products. COS-7 cells were transfected with constructs A and B. In both A and B constructs, the ATG1039 was mutated into a noninitiation codon. The proteins produced in the cell media were analyzed by Western blotting using the CAT-Ab or the N-Ab. The name of the transfected construct is indicated below the lane. B, Immunocytolocalization of the L-VEGF isoform and its products in transfected COS-7 cells using the N-Ab (green labeling). Transfected COS-7 were also treated with a commercial antibody raised against the 58K Golgi protein (red labeling) or a reticulum endoplasmic tracer (blue labeling). The constructs transfected in COS-7 cells are drawn above the pictures. Construct 1 (left panel) corresponds to the full-length VEGF cDNA in which the ATG1039 codon was mutated into noninitiating codons Lower picture represent an enlargement of a part of the upper picture (original magnifications, ×100 and ×40, respectively). In construct 2 (central panels), only nt 499–1,150 of the 5′-UTR were present, and the CTG499 was mutated into an ATG codon. Original magnification for pictures 1 and 2 was performed with ×40 objective, whereas pictures 3 and 4 were obtained using a confocal laser microscope with a ×100 objective.

The AUG 1,039-initiated form (lane A) produced from construct A and the COOH-terminal product of the cleavage of L-VEGF (lane B) generated by construct B (in the absence of AUG initiation) were both clearly secreted. Indeed, they could be detected in the culture medium by Western blotting using the CAT-Ab. In contrast, neither the L-VEGF nor its NH2 fragment was exported from the cells (lanes 1–4) as indicated by the absence of detection of the 50-kDa protein and the 23-kDa product in the culture media with the CAT-Ab or the N-Ab.

In an attempt to identify the intracellular localization of the L-VEGF and its NH2 fragment, immunocytodetections were carried out using the N-Ab on transfected COS-7 cells. Cells were transfected with a low amount of DNA and harvested 24 h later to prevent any overexpression of proteins that might perturb their localization.

Figure 8B shows that comparable staining was obtained in cells transfected with constructs that generate the full-length L-VEGF (construct 1) or only the NH2-terminal fragment (construct 2). In both cases, the labeling was localized in the reticulum-Golgi network. Indeed the staining was very similar to that observed with an anti-Golgi antibody (red labeling) or the endoplasmic reticulum staining (blue labeling).

DISCUSSION

We demonstrate here the existence, both in vitro and in vivo, of a new L-VEGF isoform that is initiated at the CUG 499 located 539 nt upstream from the classical AUG VEGF initiation codon. The 206-a.a. acid extension present in this long protein is, surprisingly, highly conserved through three mammals. We detected this isoform in various mouse tissues and also in COS-7 cells transfected with the VEGF full-length cDNA with both a specific antibody raised against this extension and a commercial VEGF antibody. In Fig. 2D, we failed to detect the classical VEGF 189 AUG-initiated protein probably due to the fact that this protein is efficiently secreted from the cells in vivo. A 49-kDa protein had already been detected in extracts prepared from epididymis using a VEGF antibody (33) under reducing conditions, although nothing was known about its origin. In this paper we propose, for the first time, a characterization of such a long VEGF isoform.

One interesting feature of the L-VEGF, revealed by our data, is its partial cleavage into two long fragments with distinct cellular localizations, namely a 206-a.a. N-terminal fragment that remains in the cell but appears to be associated with the secretion apparatus and a C-terminal fragment that is secreted. Clearly, the L-VEGF protein represents an unusual model of cleavage and protein export. This atypical internal cleavage indeed occurs in the vicinity of a hydrophobic peptide located in the central part of the L-VEGF, which plays the role of signal peptide in the classical AUG-initiated form. Point mutation of this sequence affects its signal peptide function in the AUG 1,039-initiated form and also prevents the cleavage of the L-VEGF. Obviously, the large size and the primary structure of the N-terminal sequence located upstream from the hydrophobic signal peptide in L-VEGF prevent neither Golgi membrane localization nor cleavage. It should be noted that such an unusual cleavage at the internal signal peptide level has also been described for another growth factor, the FGF-3 CUG-initiated protein. In that case, however, the position of the signal peptide is quite different since it is only 10 a.a. away from the N-terminal end of the protein (26, 34).

Another prerequisite of the L-VEGF cleavage is translocation of the protein into the endoplasmic reticulum, as shown by the data obtained in a rabbit reticulocyte lysate complemented with microsomal membranes. It should be noted that the addition of microsomes after translation completion does not induce cleavage (not shown), which thus seems to be a cotranslational process. The localization of the L-VEGF in the Golgi is not surprising. This situation is observed with the precursor proteins of TGFβ-1 or insulin, which also accumulate as immature isoforms in the Golgi structure (35, 36).

The C-terminal fragment of the L-VEGF is likely to be similar, if not identical, to the AUG-initiated VEGF. The size and posttranslational modifications appear to be identical, and both molecules exhibit the same mitogenic activity (not shown).

The function of the NH2 fragment generated by the cleavage is intriguing. So far, the screening of different protein data bases did not lead to the identification of any known structural motif in the NH2 fragment of the L-VEGF, and the possible function of this protein remains to be elucidated.

It is tempting to postulate that this fragment has a specific function since its length and primary structure are remarkably conserved throughout the three species from which a sequence is available. This contrasts, for instance, with the amino-terminal parts of the CUG-initiated FGF-2 isoforms, which are much less conserved throughout mammals.

It should be noted here that the high degree of conservation of the a.a. sequence of this NH2 fragment could, at least partially, result from the structural constraints imposed on the mRNA by the existence of IRES A. Indeed, this ribosomal entry site constitutes more than half of the NH2 extension coding sequence.

The L-VEGF protein is initiated at the level of a relatively “weak” CUG initiation codon, whereas the secreted isoform is initiated at the level of an AUG codon which is in a very poor Kozak context (37). Maybe the selection of such a CUG limits the initiation of an isoform whose expression needs to be strictly balanced with that of the AUG-initiated form to promote the VEGF action. Several growth factors (FGF-2, FGF-3) and c-myc also exist as CUG-initiated isoforms, and their physiological roles are often different from those of the AUG-initiated forms (23, 27, 38–40). Nevertheless, in the case of FGF-2, FGF-3, and c-myc, the presence of a CUG start codon is explained by the need to allow leaky scanning of the 43S ribosomal subunit that leads to recognition of the downstream AUG. In the case of VEGF, the initiation at the AUG does not require a leaky scanning mechanism since it is driven by a ribosomal entry site, IRES A.

We show that IRES B controls the translation of the L-VEGF isoform, which is cleaved to generate a COOH fragment indistinguishable from the genuine AUG 1,039-initiated form. The presence of IRES B thus appears to enhance the level of production of this protein. This explains our previous data (16), which indicated that the AUG-initiated protein was more abundantly produced when the messenger contained both IRES A and B than when only the IRES A was present.

These data also suggest that the regular VEGF protein is, in fact, produced either by initiation at the classical initiation at the AUG 1,039 or through synthesis of the L-VEGF followed by the cleavage of the NH2 fragment. The choice between these two possibilities may be determined by the cellular environment and hence the relative efficiency of the two IRES A and B. We previously demonstrated that different cellular proteins were able to bind to the two IRES A and B (16). These two ribosomal entry sites could thus act as molecular sensors of cell physiology. They may control a potential differential regulation of the alternative initiation that leads to a balanced synthesis of the L-VEGF and the AUG 1,039 VEGF isoforms.

MATERIALS AND METHODS

Molecular Cloning of the Mouse VEGF cDNA

A few grams of mouse lung were crushed with a potter in Tri-Reagent. Total RNA were prepared using the Tri-Reagent method (Euromedex, Mundolsheim, France), derived from the guanidinium thiocyanate procedure (41). RNA was extracted after chloroform addition and precipitated with isopropanol. After an ethanol wash and precipitation, the RNA was quantified by measuring the absorbance at 260 nm, and its quality was checked by electrophoresis on agarose gel with ethidium bromide staining. Reverse transcription was performed using a polydT primer as described by the manufacturer (SuperScript preamplification system, Life Technologies, Inc., Gaithersburg, MD). The cDNA were then amplified by PCR using the primers VIS-5 5′-GCTCTAGAUGCTTTTGGGGGTGACCGCC-3′ or VIS2–5 5′-GCTCTAGACAAGAGCTCCAGAGAGAAGTC-3′ (sense primers), and VIS2–3 5′-TTTATCGATCUGGGACCACTTGGCAUGG-3′ (antisense). The products of three independent PCRs were digested by _Cla_I and _Xba_I and cloned into _Xba_I-_Cla_I-digested PKS vector. The cloned cDNAs were then sequenced in both directions using T7 and T3 primers. The VEGF mouse 5′-UTR sequence that we had obtained was compared with that of the corresponding human and bovine sequences using CLUSTAL software (Infobiogen, Evry, France).

Plasmid Constructions

The human VEGF 189 coding sequence and the DNA fragment corresponding to the 5′-UTR of the messenger were kindly provided by J. Abraham (29). These fragments were cloned into pSCT-derived plasmid (42) downstream from the cytomegalovirus and T7 promoters between the _Xba_I and _Bgl_II sites to generate the p5′189 plasmid (Fig. 2A). The p189 plasmid was obtained by deletion of the 5′-UTR (up to _Nae_I site in position 1,012) from p5′189. The construct of the plasmid pVC (C in Fig. 3A) has already been described (16). The pVC1012 construct (A in Fig. 3A) was obtained after deletion of the 5′-UTR up to the _Nae_I site (position 1,012), from the pVC plasmid. The pVC585 (B in Fig. 3A) results from the deletion of the _Xba_I-_Sac_II (nt 585) fragment from the pVC. The pVC-derived pVCATG vector (D in Fig. 3A) contains the entire 5′-UTR and the mutation of the CTG codon (position 499) into an ATG codon. This construct is the result of the cloning into the _Pvu_II and _Xba_I sites of pVC (positions 483 and 917 in the 5′-UTR, respectively) of the _Pvu_II-_Xba_I-digested PCR fragment obtained after amplification of the 5′-UTR with the primers +1CTG 5′-TGGGATCCCGCAGCTGACCAGTCGCC ATGGCG-3′ and VN3′. The pVCTTT vector (E in Fig. 3A) contains the full 5′-UTR in which the CTG codon in position 499 was mutated into a TTT codon. This mutation was generated using the primer +1TTT 5′-TTGGGATCCCGCAGCTGACCAGTCGCG TTTCCGGACA-3′ and VN3′ 5′-AAAGGAGCTCAGATCTATTAGGTTTCGGAGGCCCGACC-3′. The amplified fragment of 5′-UTR was digested and cloned into the _Bam_HI and the _Sma_I sites of the pVC vector.

To introduce the mutation of the ATG1039 in the pVC1012, pVC, and pVCATG, the following plasmids were constructed.

The pVC1012CCC (construct 2 in Fig. 5B) harbors the mutation of the ATG1039 into a CCC codon. It was obtained by cloning a PCR product digested _Ngo_MI kle-_Sac_I into the _Xba_I kle-_Sac_I sites of the PSCT12VCAT vector (42). The PCR fragment was amplified from the pVC plasmid using the primers ATG-CCC 5′: CCGCGCCGGCCCCGGTCGGGCCTCCGAAACC CCCAACTTTCTG-3′ (matching residues 1,008 to 1,044) and “ 4REV, ” complementary to the end of the CAT gene, whose sequence is 5′-TTTGAGCTCAGATCTCATTACGCCCCGCCCTGCCA-3′. The pVCCCC plasmid (construct 4 in Fig. 5B) was obtained by cloning the same PCR fragment into the pVC vector digested at the _Ngo_MI (nt 1,012) and _Sac_I sites (end of the CAT gene). The mutation of the ATG codon 1039 was also introduced into the pVCATG vector by inserting the _Nhe_I-_Xho_I fragment from PVCCCC in place of the _Nhe_I-_Xho_I fragment of pVCATG to form the pVCATG-CCC vector (construct 6 in Fig. 5B). The pRVCTG plasmid (construct 3 in Fig. 5A) is a pRF11AEN-derived vector (27) in which a _Xba_I-_Sac_I-digested PCR fragment obtained after amplification of the 5′-UTR with the primers VN5′ 5′-AAATCTAGACTCGAGACC ATGGGAACGGACAGACAGACAGAC-3′ and VN3′ was cloned into the _Xba_I-_Sac_I-digested pRF11AEN. This vector contains nt 499–1,038 of the 5′-UTR and the mutation of the CTG (position 499) into the ATG codon without any CAT coding sequence. The PVSP plasmid (construct 4 in Fig. 5A) was obtained by cloning a PCR fragment into the _Nhe_I and the _Bgl_II sites of the PVCATG499 vector (see below). This fragment (nt 736–1,117) resulted from the amplification of PVC with the oligonucleotides NHE and PSSTOPAS 5′-AAAAGATCT ATTAAGCCTGGGACCACTTGGCATG-3′. In this way, two stop codons were created just downstream from the VEGF signal peptide coding sequence (nt 1,039–1,115).

The mutation of the signal peptide was performed using the hybridizing PCR method. A first PCR fragment was obtained after amplification of the pVC plasmid with the primers PSS (matching nt 1,059–1,084) 5′-GGTG GATTGGAGCCTTGCC GAGCTGC-3′ and 4REV (matching the end of the CAT gene). This PCR product was hybridized with a second amplification product obtained from pVC with the primers NHE (matching nt 736–754) and PSAS, complementary with the PSS, 5′-GCAGCTCGGCAAGGCTCCAATCCACC-3′. This hybrid served as template for the third amplification using primers NHE and 4REV. The product was then digested at the _Xba_I and _Dra_III sites (positions 917 and 1,150 on the amplified fragment, respectively) and cloned into the _Xba_I-_Dra_III-digested pVC and pVCATG-CCC vectors to create the plasmids pVCmSP (construct A′ in Fig. 6A) and pVCATG-CCCmSP (construct 2 in Fig. 6B).

The pVC499ATG (construct 1 in Fig. 8B) is a pVC-derived plasmid in which the 5′-UTR has been deleted between nt 1 and 499 and the CTG codon in position 499 mutated into an ATG codon. For technical convenience the ATG was designed to become an _Nco_I site, which led to the addition of a GGA codon downstream from the ATG. To obtain this plasmid, a PCR was performed using the primers VN5′ and VN3′. The amplified product extending from positions 499 to 1,038 was digested with _Xba_I and _Sma_I and cloned into the pVC plasmid. The primer generated this _Xba_I site. The construction of the pCVC and pHCVC vectors (constructs A and B in Fig. 4) has been described previously (16).

In Vitro Translation And Immunoprecipitation

The plasmids 5′189 and 189 were linearized downstream from the 3′-end of the VEGF coding sequence at the _Bgl_II site. Uncapped mRNA was generated in vitro using the T7 m-message machine kit (Ambion, Inc., Austin, TX) according to manufacturer’s instructions. mRNA was quantified by absorbance at 260 nm. Ethidium bromide staining of agarose gel was used as a quality control. In vitro translation in rabbit reticulocyte lysate (Promega Corp., Madison, WI) was performed as previously described (42), in the presence of 35S-methionine (Amersham Pharmacia Biotech, Piscataway, NJ). For the microsome-induced maturation assay, 35S-methionine (ICN, Orsay, France) was used to label the translation products and 1 μl of canine pancreatic microsomal membranes (Promega Corp., Roche, Indianapolis, IN) were added per 10 μl of lysate. Seven microliters (or 10 μl when microsomes were added) of translation sample were analyzed by electrophoresis in a 12.5% SDS-polyacrylamide gel. The neosynthesized proteins in Fig. 2A were immunoprecipitated with pansorbin as previously described: 20 μl of the translation sample were diluted to 150 μl in the PBS/NP40 buffer (PBS 1×, 50 mm NaF, 2 mm EDTA, 2 mm EGTA, 0.05% NP40) and precleared by incubation with 50 μl of pansorbin for 10 min at room temperature. The supernatant was incubated for 30 min at room temperature with 10μ l of N-Ab or VEGF-Ab (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, sc-152) and then for 30 min at room temperature with 50μ l of pansorbin. After five washes in HEPES/NP40 buffer (150 mm NaCl, 15 mm HEPES, pH 7.4, 1 mm EDTA, pH 7.4, 0.5% NP40), the samples were analyzed by 12.5% PAGE as above.

DNA Transfection and Western Blotting

COS-7 monkey cells were transfected using Fugene-6 transfection reagent (Roche) according to manufacturer’s instructions. In the case of tunicamycin treatment, cells were incubated 12 h after transfection with 100 μm tunicamycin (Roche) for 24 h. Twenty-four to 48 h after transfection, cells were lysed in dithiothreitol andβ -mercaptoethanol-containing sample buffer (1×) and scraped. The samples were heated for 2 min at 95 C. For analysis of secreted proteins, transfections were performed in the absence of serum, and proteins from the cell media were precipitated with 7 volumes of acetone or 20% of acetic acid. The protein pellets were then resuspended in equal volumes of sample buffer (1×). The proteins were separated in a 12.5% polyacrylamide gel and transferred onto a nitrocellulose membrane. CAT proteins were immunodetected using rabbit polyclonal CAT-Ab prepared in the laboratory (1:10,000 dilution). The VEGF proteins were detected using VEGF-Ab (Santa Cruz Biotechnology, Inc., sc-152) (dilution 1:300). The N-Ab was prepared in the laboratory. The antigen injected into the rabbits corresponds to the a.a. sequence deduced from the cDNA sequence located between nt 554 (_Nae_I site) and 937 (_Sma_I site). Antibodies were detected using an enhanced chemiluminescence kit (Amersham Pharmacia Biotech).

Immunocytolocalization

COS-7 cells were seeded on lamella before transfection. Twenty-four hours after fugene-6 transfection, the cells were washed four times with PBS and fixed for 7 min. in 3% paraformaldehyde. They were then washed again three times for 5 min each, twice with PBS and once with PBS plus NH4Cl 50 mm, and permeabilized for 20 min at 37 C with nonimmune antibody (1:100 in PBS-0.025% saponin-0.5% BSA). The cells were then incubated with the N-Ab (1:200) or the anti-58K antibody (1:200, Sigma, St. Louis, MO; G2404) in PBS-0.025% saponin-0.5% BSA, respectively) for 90 min at 37 C, washed again three times with PBS-0.025% saponin, and incubated for 40 min at room temperature with fluorescein isothiocyanate-coupled antirabbit antibody (1:400, Tebu, Le Perray, France) and washed again three times. The lamella was fixed on microscope slides with Citifluor (Pelco, Inc.). Labeling of the endoplasmic reticulum in Fig. 8B was obtained by incubating fixed cells with 1 μm ER tracker blue-white DPX (Molecular Probes, Inc., Eugene, OR) during 30 min following manufacturer’s instructions.

Tissue Protein Extraction

Tissues were crushed with a potter in a buffer made of Tris-HCl, pH 7.5, 50 mm, NaCl 150 mm, EDTA 1 mm, in the presence of a protease inhibitors mix. The lysate was filtered and centrifuged at 105,000 × g for 30 min. The supernatant corresponding to the cytoplasmic fraction was aliquoted. The pellet was then resuspended in a buffer made of Tris-HCl, pH 7.5, NaCl 150 mm, NP 40 1%, SDS 0.1%, and sodium deoxycholate 0.5%. The suspension was sonicated and recentrifuged for 20 min at 1,000 × g. The pellet was discarded and the supernatant was aliquoted. The protein amount was quantified using the BCA protein assay (Pierce Chemical Co., Rockford, IL).

Acknowledgments

We thank Y. Audigier, F. Bayard, J. Plouët, S. Audigier, and A. C. Prats for helpful discussions.

This work was supported by grants from the Ligue Nationale contre le Cancer. I.H. received a fellowship from the Association pour la Recherche contre le Cancer. S.B. received a fellowship from the Ministère de l’Education Nationale et de la Recherche Scientifique. L.C. received a fellowship from the Aventis Company.

Abbreviations:

a.a.,
CAT,
chloramphenicol acetyltransferase;
CAT-Ab,
FGF,
fibroblast growth factor;
IRES,
internal ribosome entry site;
L-VEGF,
N-Ab,
nt,
ORF,
RRL,
rabbit reticulocyte lysates;
UTR,
VEGF,
vascular endothelial growth factor;
VEGF-Ab,

Ferrara

1999

Molecular and biological properties of vascular endothelial growth factor.

J Mol Med

527

–

543

Perrella

Tsai

Yet

Hsieh

Yoshizumi

Patterson

Endege

Zhou

Lee

1995

Induction of vascular endothelial growth factor gene expression by interleukin-1β in rat aortic smooth muscle cells.

J Biol Chem

270

308

–

312

Cohen

Nahari

Cerem

Neufeld

Levi

1996

Interleukin 6 induces the expression of vascular endothelial growth factor.

J Biol Chem

271

736

–

741

Goad

Rubin

Wang

, Tashjian

Patterson

1996

Enhanced expression of vascular endothelial growth factor in human SaOS-2 osteoblast-like cells and murine osteoblasts induced by insulin-like growth factor I.

Endocrinology

137

2262

–

2268

Brogi

Namiki

Isner

1994

Indirect angiogenic cytokines upregulate VEGF and bFGF gene expression in vascular smooth muscle cells, whereas hypoxia upregulates VEGF expression only.

Circulation

649

–

652

Mukhopadhyay

Tsiokas

Zhou

Foster

Brugge

Sukhatme

1995

Hypoxic induction of human vascular endothelial growth factor expression through c-Src activation.

Nature

375

577

–

581

Grugel

Finkenzeller

Weindel

Barleon

Marme

1995

Both v-Ha-Ras and v-Raf stimulate expression of the vascular endothelial growth factor in NIH 3T3 cells.

J Biol Chem

270

25915

–

25919

Rak

Mitsuhashi

Bayko

Filmus

Shirasawa

Sasazuki

Kerbel

1995

Mutant ras oncogenes upregulate VEGF/VPF expression: implications for induction and inhibition of tumor angiogenesis.

Cancer Res

4575

–

4580

Shweiki

Itin

Soffer

Keshet

1992

Vascular endothelial growth factor induced by hypoxia may mediate hypoxia-initiated angiogenesis.

Nature

359

843

–

845

Houck

Leung

Rowland

Winer

Ferrara

1992

Dual regulation of vascular endothelial growth factor bioavailability by genetic and proteolytic mechanisms.

J Biol Chem

267

26031

–

26037

Poltorak

Cohen

Sivan

Kandelis

Spira

Vlodavsky

Keshet

Neufeld

1997

VEGF145, a secreted vascular endothelial growth factor isoform that binds to extracellular matrix.

J Biol Chem

272

7151

–

7158

Stein

Neeman

Shweiki

Itin

Keshet

1995

Stabilization of vascular endothelial growth factor mRNA by hypoxia and hypoglycemia and coregulation with other ischemia-induced genes.

Mol Cell Biol

5363

–

5368

Claffey

Senger

Spiegelman

1995

Structural requirements for dimerization, glycosylation, secretion, and biological function of VPF/VEGF.

Biochim Biophys Acta

1246

–

Plouet

Moro

Bertagnolli

Coldeboeuf

Mazarguil

Clamens

Bayard

1997

Extracellular cleavage of the vascular endothelial growth factor 189-amino acid form by urokinase is required for its mitogenic effect.

J Biol Chem

272

13390

–

13396

Houck

Ferrara

Winer

Cachianes

Leung

1991

The vascular endothelial growth factor family: identification of a fourth molecular species and characterization of alternative splicing of RNA.

Mol Endocrinol

1806

–

1814

Huez

Creancier

Audigier

Gensac

Prats

1998

Two independent internal ribosome entry sites are involved in translation initiation of vascular endothelial growth factor mRNA.

Mol Cell Biol

6178

–

6190

Stein

Itin

Einat

Skaliter

Grossman

Keshet

1998

Translation of vascular endothelial growth factor mRNA by internal ribosome entry: implications for translation under hypoxia.

Mol Cell Biol

3112

–

3119

Akiri

Nahari

Finkelstein

Elroy-Stein

Levi

1998

Regulation of vascular endothelial growth factor (VEGF) expression is mediated by internal initiation of translation and alternative initiation of transcription.

Oncogene

227

–

236

Miller

Dibbens

Damert

Risau

Vadas

Goodall

1998

The vascular endothelial growth factor mRNA contains an internal ribosome entry site.

FEBS Lett

434

417

–

420

Bernstein

Sella

Elroy-Stein

1997

PDGF2/c-sis mRNA leader contains a differentiation-linked internal ribosomal entry site (D-IRES).

J Biol Chem

272

9356

–

9362

Vagner

Gensac

Maret

Bayard

Amalric

Prats

1995

Alternative translation of human fibroblast growth factor 2 mRNA occurs by internal entry of ribosomes.

Mol Cell Biol

–

Teerink

Voorma

Thomas

1995

The human insulin-like growth factor II leader 1 contains an internal ribosomal entry site.

Biochim Biophys Acta

1264

403

–

408

Prats

Kaghad

Prats

Klagsbrun

Lelias

Liauzun

Chalon

Tauber

Amalric

Smith

Caput

1989

High molecular mass forms of basic fibroblast growth factor are initiated by alternative CUG codons.

Proc Natl Acad Sci USA

1836

–

1840

Hann

Sloan-Brown

Spotts

1992

Translational activation of the non-AUG-initiated c-myc 1 protein at high cell densities due to methionine deprivation.

Genes Dev

1229

–

1240

Barraille

Chinestra

Bayard

Faye

1999

Alternative initiation of translation accounts for a 67/45 kDa dimorphism of the human estrogen receptor ERα.

Biochem Biophys Res Commun

257

–

Acland

Dixon

Peters

Dickson

1990

Subcellular fate of the int-2 oncoprotein is determined by choice of initiation codon.

Nature

343

662

–

665

Arnaud

Touriol

Boutonnet

Gensac

Vagner

Prats

1999

A new 34-kilodalton isoform of human fibroblast growth factor 2 is cap dependently synthesized by using a non-AUG start codon and behaves as a survival factor.

Mol Cell Biol

505

–

514

Leung

Cachianes

Kuang

Goeddel

Ferrara

1989

Vascular endothelial growth factor is a secreted angiogenic mitogen.

Science

246

1306

–

1309

Tischer

Mitchell

Hartman

Silva

Gospodarowicz

Fiddes

Abraham

1991

The human gene for vascular endothelial growth factor. Multiple protein forms are encoded through alternative exon splicing.

J Biol Chem

266

11947

–

11954

Shima

Kuroki

Deutsch

Adamis

D’Amore

1996

The mouse gene for vascular endothelial growth factor. Genomic structure, definition of the transcriptional unit, and characterization of transcriptional and post-transcriptional regulatory sequences.

J Biol Chem

271

3877

–

3883

Levy

Wegner

Goldberg

1995

Transcriptional regulation of the rat vascular endothelial growth factor gene by hypoxia.

J Biol Chem

270

13333

–

13340

Fuxe

Raschperger

Pettersson

2000

Translation of p15.5INK4B, an N-terminally extended and fully active form of p15INK4B, is initiated from an upstream GUG codon.

Oncogene

1724

–

1728

Ergun

Empen

Fiedler

1997

VEGF modulates the capillaries of the human epididymis.

Adv Exp Med Biol

424

189

–

190

Kiefer

Acland

Pappin

Peters

Dickson

1994

Competition between nuclear localization and secretory signals determines the subcellular fate of a single CUG-initiated form of FGF3.

EMBO J

4126

–

4136

Kjeldsen

Pettersson

Hach

1999

The role of leaders in intracellular transport and secretion of the insulin precursor in the yeast Saccharomyces cerevisiae.

J Biotechnol

195

–

208

Miyazono

Thyberg

Heldin

1992

Retention of the transforming growth factor-β 1 precursor in the Golgi complex in a latent endoglycosidase H-sensitive form.

J Biol Chem

267

5668

–

5675

Kozak

1989

The scanning model for translation: an update.

J Cell Biol

108

229

–

241

Hann

Dixit

Sears

Sealy

1994

The alternatively initiated c-Myc proteins differentially regulate transcription through a noncanonical DNA-binding site.

Genes Dev

2441

–

2452

Creancier

Morello

Mercier

Prats

2000

Fibroblast growth factor 2 internal ribosome entry site (IRES) activity ex vivo and in transgenic mice reveals a stringent tissue-specific regulation.

J Cell Biol

150

275

–

281

Creancier

Mercier

Prats

Morello

2001

c-myc Internal ribosome entry site activity is developmentally controlled and subjected to a strong translational repression in adult transgenic mice.

Mol Cell Biol

1833

–

1840

Chomczynski

Sacchi

1987

Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction.

Anal Biochem

162

156

–

159

Prats

Vagner

Prats

Amalric

1992

cis-Acting elements involved in the alternative translation initiation process of human basic fibroblast growth factor mRNA.

Mol Cell Biol

4796

–

4805