VEGF₁₆₅b, an Inhibitory Splice Variant of Vascular Endothelial Growth Factor, Is Down-Regulated in Renal Cell Carcinoma¹

Tumor survival, growth, and subsequent metastasis are underpinned by the creation of new blood vessels. The physiological or pathological formation of new vessels is a complex process controlled by at least 10 endothelium-specific growth factors from the VEGF,³ angiopoietin, and ephrin families. Other nonendothelial cell-specific growth factors, molecules, and enzymes also play a role, including cytokines, proteinases, adhesion, and junctional molecules. The terms vasculogenesis, angiogenic remodeling, and angiogenic sprouting refer to specific stages of new vessel formation, and each appears to be stimulated by a specific pattern of molecular interactions (reviewed in Refs. 1, 2).

Although the exact orchestration of tumor vascularization is still debated, VEGF-A seems to be the predominant vascular growth factor in most tumors. VEGF-A inhibition has therefore received the most attention in relation to potential anti-angiogenesis/anti-tumor therapy. Exon splicing of the VEGF pre-RNA results in three main mRNA species that code for three secreted isoforms, VEGF₁₈₉, VEGF₁₆₅, and VEGF₁₂₁ (3). A number of minor splice variants, VEGF₂₀₆, VEGF₁₈₃, VEGF₁₄₅ and VEGF₁₄₈, have been reported, but their significance remains uncertain (4, 5, 6, 7).

An increase in VEGF mRNA expression has been identified in almost all known tumors (3). A positive correlation between tumor VEGF-A expression and tumor vascularity has been described, and in many studies a correlation with prognosis has been shown (8). This is also true in RCC. Increased RCC VEGF-A expression has been shown by semiquantitative RT-PCR (9), competitive quantitative RT-PCR (10), in situ hybridization (11), and Western blotting [3–37-fold increase compared with normal tissue (9)]. These reports were recently underpinned by a quantitative “real-time” RT-PCR study based on the use of fluorogenic probes that suggested that VEGF mRNA in RCC was 200-2000 times higher than in adjacent normal tissue (12). This increase appears be true for all three common VEGF isoforms (121, 165, and 189). However, one study suggested that not all RCCs express all three common isoforms, and that the presence of VEGF₁₈₉ was a bad prognostic sign being associated with the higher pathological grades (13). This contrasts with our experience in which we find all three common isoforms in United Kingdom nephrectomy specimens (7). Similar to other tumors, VEGF expression in conventional RCCs has been shown to correlate with tumor vascularity and in addition be a significant predictor of outcome (14). Specific VEGF inhibitors have been shown to inhibit renal tumor growth and metastasis in animal models (15).

We have reported previously the identification of VEGF₁₄₈ mRNA in a study of VEGF-A isoform expression in single glomeruli in which the terminal half of exon 7 is spliced out (7). If translated, this mRNA species would code for a truncated form of VEGF₁₆₅, because a frameshift prevents translation of exon 8 by insertion of a stop codon. Because the COOH terminal of VEGF-A appears to be essential for mitogenesis and VEGF-A exerts its effects as a dimer (16), we hypothesized that VEGF₁₄₈ may act as a native inhibitor of angiogenesis through the formation of heterodimers, as has been described for other dimerized molecules, e.g., p73 (17).

We were therefore prompted to study the expression of VEGF₁₄₈ in renal cell carcinomas and normal renal tissue from the same kidney, the polar nature of renal cell carcinomas lending itself well to the comparison of normal and malignant tissue. During this study, we identified a previously undescribed isoform of VEGF, VEGF₁₆₅b, and investigated its expression characteristics and functional properties.

Non-Von Hippel Lindau nephrectomy tissue was derived from patients undergoing nephrectomy for unipolar renal tumor (age range, 47–75 years). All patients were non-diabetic and normotensive with normal excretory renal function and no urinary sediment. The non-affected renal tissue was normal on H&E staining. Immediately after nephrectomy 1 × 1 × 1-cm cubes of non-necrotic tissue were taken from the macroscopic tumor of all nephrectomy specimens. Similar cubes of normal tissue were collected from the most distant region of the opposite pole. An example of a kidney with a RCC is shown in Fig. 1. Examples of areas of tissue sampled are shown. Non-kidney tissue was obtained from postmortem specimens, donor patients, and from births. All samples were taken from cadavers within a few hours of death during solid organ retrieval from donors, or immediately after birth, respectively. Consent was obtained from patients or relatives as appropriate. Tissues were taken with local Ethical Committee approval.

All tumors were extensively sampled and reported by an experienced consultant pathologist using the Tumor-Node-Metastasis classification of malignant tumors (UICC, 1997; Ref. 18) with nuclear grading using the Fuhrman grades 1–4 (19).

One hundred to 200 mg of tissue were homogenized in Trizol reagent, and mRNA was extracted using the method of Chomczynski and Sacchi (20). Eight % of the mRNA was reverse transcribed using Moloney murine leukemia virus reverse transcriptase and poly(dT) as a primer. The cDNA was then amplified using primers designed to detect VEGF₁₄₈ (Fig. 2). One μm of a primer complementary to the 3′ UTR primer (3′ UTR, 5′-ATG GAT CCG TAT CAG TCT TTC CT-3′) and 1 μm of primers complementary to exon 7a (Exon7a, 5′-GTA AGC TTG TAC AAG ATC CGC AGA CG-3′), and 1.2 mm MgCl₂, 200 μm deoxynucleotide triphosphates, and 1 unit of Taq polymerase (Abgene) were used in reactions cycled 35 times, denaturing at 96°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 60 s. PCR products were run on 3% agarose gels containing 0.5 μg/ml ethidium bromide and visualized under a UV transilluminator. This reaction consistently resulted in one amplicon at ∼150 bp (consistent with VEGF₁₄₈), and one amplicon at ∼200 bp, consistent with VEGF₁₆₅ VEGF₁₈₃, VEGF₁₈₉, and VEGF₂₀₆.

The band at ∼150 bp was excised from the gel under UV transillumination, and the DNA was extracted using Qiaex (Qiagen). The DNA was then digested with appropriate enzymes and ligated into pBluescriptKSII (Stratagene). Ligations were transformed into supercompetent XL-1 Blue Escherichia coli (Stratagene) and grown on ampicillin-resistant Luria broth agar plates. Colonies were amplified, and the plasmid DNA was purified using Qiagen columns. The DNA was then sequenced using T7 and T3 sequencing primers by fluorescent dideoxy termination sequencing (ABI370). Sequences were analyzed by automated fluorescent chromatography, and the sequence was confirmed against the chromatograph manually.

In four of the samples, the normal tissue had significantly more ∼150-bp product than ∼200-bp product. One of these samples was used to confirm the full length of the new isoform. The entire length of the product was amplified by RT-PCR using primers downstream of the original 3′ UTR primer (V165X, 5′-AAT CTA GAC CTC TTC CTT CAT TTC AGG-3′) engineered with an XbaI site, and a primer complementary to the translation initiation site of the other isoforms of VEGF (V165K, 5′-CCG GTA CCC CAT GAA CTT TCT GC-3′) engineered with a KpnI site and PCR conditions as described above. This resulted in production of amplicons of a variety of lengths (as described in Table 1). These were digested with KpnI and XbaI and ligated into pBluescriptKSII (Stratagene). These were then screened by PCR using exon 7 and 3′ UTR primers to reveal colonies that contained the novel splice variants.

cDNA was amplified using primers designed to detect exon 9-containing isoforms. One μm of a primer complementary to the last five bases of exon 7 and the whole of exon 9 (Exon 9, 5′-TCA GTC TTT CCT GGT GAG AGA TCT GCA-3′), 1 μm of primers complementary to exon 4 (Exon 4, 5′-GAG ATG AGC TTC CTA CAG CAC-3′), 1.2 mm MgCl₂, 200 μm deoxynucleotide triphosphates, and 1 unit of Taq polymerase (Abgene) were used in reactions cycled 35 times, denaturing at 96°C for 30 s, annealing at 63°C for 30 s, and extension at 72°C for 60 s. PCR products were run on 3% agarose gels containing 0.5 μg/ml ethidium bromide and visualized under a UV transilluminator. To confirm specificity of the primers, increasing concentrations of VEGF₁₆₅b were mixed with increasing concentrations of VEGF₁₆₅ cDNA, and PCR was carried out using exon 9 and exon 4 primers (Fig. 3 a).

Full-length VEGF₁₆₅b or VEGF₁₆₅ (generated by PCR from tissue) was cloned into the expression vector pcDNA₃ using standard methodology and then transfected into HEK 293 cells, and a stable cell line was generated using a Geneticin selection. Confluent cells were incubated in basal M200 endothelial cell medium for 48 h, containing neither serum nor Geneticin, and the conditioned medium was assayed for VEGF concentration using a pan VEGF ELISA (R&D). Control conditioned medium (pcDNA3-CM) was collected in the same way from HEK293 cells stably transfected with pcDNA₃. Fig. 3 b shows a Western blot of conditioned medium from VEGF₁₆₅b-transfected cells (VEGF₁₆₅b-CM) and VEGF₁₆₅-transfected cells (VEGF₁₆₅-CM). VEGF₁₆₅b was the same molecular weight as VEGF₁₆₅, and both isoforms correspond to previously published molecular weights for VEGF₁₆₅b (21). The blot also confirms that most of the VEGF₁₆₅ and VEGF₁₆₅b made in HEK cells had a slightly greater molecular weight than commercially available VEGF₁₆₅ (M_r 23kDa compared with M_r 18kDa). This is probably because commercially available VEGF₁₆₅ is not glycosylated. There was some VEGF₁₆₅b and VEGF₁₆₅ in the conditioned medium that appeared to be deglycosylated, but this was a small fraction of the total VEGF.

Freshly isolated HUVECs incubated in 0.1% serum M200 overnight were incubated with either pcDNA3-CM, VEGF₁₆₅-CM (adjusted to 100 ng/ml VEGF₁₆₅ with pcDNA3-CM), VEGF₁₆₅b-CM (adjusted to 100 ng/ml VEGF₁₆₅b with pcDNA3-CM), or a mixture of VEGF₁₆₅b-CM and VEGF₁₆₅-CM (adjusted to 100 ng/ml VEGF₁₆₅b and 100 ng/ml VEGF₁₆₅ with pcDNA3-CM), and then 37 kBq of [³H]thymidine (Amersham Pharmacia) added. After 4 h, the cells were washed and trypsinized, cell number was counted on a hemocytometer, and radioactivity was measured on a beta counter (LKB-1217). Incorporation was calculated as counts/cell. Dose-response curves were carried out in a similar manner, except that VEGF₁₆₅b-CM was diluted with pcDNA3-CM, and 50 ng/ml VEGF₁₆₅ (Peprotech, NJ) added. Proliferation index was calculated as [³H]thymidine incorporation of VEGF₁₆₅-treated cells compared with mean incorporation into cells with no VEGF₁₆₅ treatment.

Migration assays were performed in a modified 24-well Boyden chamber containing collagen-coated polycarbonate filter inserts (8 μm pore; Millipore). The filters were placed in 24-well plates containing 0.5 ml/well of either: (a) VEGF₁₆₅-CM containing 33 ng/ml VEGF₁₆₅; (b) VEGF₁₆₅b-CM containing 33 ng/ml VEGF₁₆₅b; (c) VEGF₁₆₅-CM and VEGF₁₆₅b-CM (33 ng/ml of each isoform); or (d) pcDNA₃-CM. HUVECs were suspended in serum-free medium, and 25,000 cells were added to the upper chamber of each well. The plate was incubated for 6 h to allow migration, medium was removed, and both chambers were washed with PBS (two times). 0.2 mg/ml thiazolyl blue (MTT) in medium was then added to both chambers and incubated for 3 h at 37°C. The medium was removed, and the chambers were washed with PBS (two times). Nonmigrated cell crystals in the upper chamber (stained blue) were removed with a cotton swab, which was placed in 1 ml of DMSO to dissolve the MTT product. Migratory cell crystals (on the underside of the insert) were also dissolved in MTT. The samples were left overnight to permit complete solution of the product. The absorbance of soluble MTT was determined at a wavelength of 570 nm using a spectrophotometer. The percentage migration was then calculated from the intensity of the lower well as a percentage of the total intensity of both wells. Assays were run in sextuplet.

Third-order superior mesenteric arteries were dissected from 200- to 300-g male Wistar rats (sacrificed by stunning and cervical dislocation), and leak-free segments were mounted in an arteriograph at 80 mm Hg, in rat Ringer. Arteries were constricted with 0.6–1μm phenylephrine applied in the superfusate. Ten μm Acetylcholine maximally dilated all arteries used for these experiments, demonstrating the presence of an intact endothelium. ACh, CM, and VEGF isoforms, dialyzed against rat Ringer’s solution using M_r 3.5kDa dialysis tubing were applied to the lumen of the artery using an adaptation of the Halpern pressure myograph technique as described previously (22). The concentration of VEGF in the CM after dialysis was determined by ELISA. All data are expressed as mean ± SE for four experiments. Statistical significance was tested using ANOVA and Student Newmann Keuls post-hoc test.

An amplicon consistent with VEGF₁₄₈ expression was generated by RT-PCR from mRNA extracted from normal human kidney samples. An example of the PCR products of four kidneys (benign and malignant tissues) is shown in Fig. 4,a. This amplicon was detected in 17 of 18 non-malignant, histologically normal samples of human kidney but in a significantly lower proportion of malignant samples (4 of 18; Fig. 4 b; P < 0.001; Fisher’s exact test). In addition, the median stage of the tumors of which PCR did not result in production of this amplicon was 3b (interquartile range, 2–4), whereas the median stage of the four tumors that did result in this amplicon after PCR was 2 (interquartile range, 1–2.25).

Subsequent sequencing of this amplicon revealed an unexpected 3′ sequence (Fig. 5). This unexpected sequence was found in all seven PCR samples sequenced. The sequence indicated that the mRNA was not VEGF₁₄₈ but was spliced from the 3′ end of exon 7 into the 3′ untranslated region of VEGF₁₆₅ mRNA 44 bp downstream of the end of exon 8 (23). This splice site has the same first two nucleotides as exon 8 but a different 3′ sequence (Fig. 5,a). We have termed this exon 9. The intronic region between exon 7 and exon 9 has an intronic consensus sequence of 5′-GT… CCAG-3′ and a high CT-rich region 6–24 bp prior to the 5′ end of exon 9 (Fig. 5,b). PCR of the full-length product using primers V165K and V165X resulted in one strong band at ∼670 bp. Furthermore, nested PCR using the 3′ UTR and 7a primers described above resulted in a strong band at ∼130 bp, confirming that the full length was VEGF₁₆₅b. The full-length product was a 663-nucleotide sequence encoding a peptide 191 amino acids long. This peptide consisted of the same NH₂-terminal 185 amino acids as VEGF₁₆₅ (i.e., a 26-amino acid signal sequence followed by 159 amino acids corresponding to exons 1, 2, 3, 4, 5, and 7). However, the COOH-terminal 6 amino acids were not the same as exon 8 (Fig. 5 c). The six amino acids that this new exon codes for are Ser-Leu-Thr-Arg-Lys-Asp (SLTRKD), followed by a stop codon, TGA. Because this splice variant would code for a mature 165-amino acid polypeptide (after cleavage of the signal sequence) with 96.4% identity with VEGF₁₆₅, we have termed this isoform VEGF₁₆₅b. This exact same COOH-terminal exon structure was seen in all clones sequenced that produced a PCR product of ∼130 bp with primers exon 7 and 3′ UTR.

mRNA from 16 different tissues was reverse transcribed and amplified using exon 7 and 3′ UTR primers. PCR products of lengths consistent with expression of exon 9 containing isoforms were clearly detected in umbilical cord, cerebrum, aorta, prostate, pituitary, lung, skeletal muscle, and placental tissue as well as kidney (Fig. 6,a). Fainter bands were also seen in colon, skin, bladder, and spinal cord. We did not detect significant exon 9 containing isoforms in hypothalamus, inferior vena cava, or liver. Subsequent PCR using exon 9-specific primers confirmed the distribution of expression in the different tissues (Fig. 6 b), but in this case, we did detect expression in liver, and expression in the pituitary was marginal. Interestingly, in aorta, prostate, and umbilical cord a slightly longer band was seen. It is not clear whether these are additional exon 9-containing isoforms such as VEGF₁₈₉b.

To determine the functional effect of VEGF₁₆₅b, we measured endothelial cell proliferation by determination of [³H]thymidine incorporation/cell number during incubation with conditioned medium from transfected cells (Fig. 7,a). The full-length cDNA generated by PCR was cloned into an expression vector (pcDNA3-VEGF₁₆₅b), as was full-length VEGF₁₆₅ (pcDNA3-VEGF₁₆₅), and each was transfected into HEK293 cells. The VEGF concentration of cell conditioned medium (VEGF₁₆₅b-CM), as determined by ELISA using pan VEGF antibodies, ranged from 80 to 400 ng/ml VEGF. CM from cells transfected with pcDNA3-VEGF₁₆₅ (VEGF₁₆₅-CM) had a VEGF concentration of 100–260 ng/ml. Medium from cells transfected with pcDNA3 alone (pcDNA3-CM) contained <62.5 pg/ml VEGF (the minimum detection limit of the ELISA). HUVECs were incubated in VEGF₁₆₅-CM adjusted to 100 ng/ml (with pcDNA3-CM), and this resulted in a significant 283 ± 43% increase in thymidine incorporation/endothelial cell compared with pcDNA3-CM alone (P < 0.01, ANOVA; Fig. 7,a). VEGF₁₆₅b-CM containing 100 ng/ml VEGF₁₆₅b did not stimulate HUVEC proliferation (165 ± 27% of pcDNA3-CM; no significant increase, but significantly less than VEGF₁₆₅-CM; P < 0.05). Furthermore, there was no increase in endothelial cell proliferation when cells were incubated in a combination of both VEGF₁₆₅b-CM and VEGF₁₆₅-CM containing 100 ng/ml of each VEGF isoform (150 ± 18% of pcDNA3-CM; not significantly different from pcDNA3-CM, but significantly lower than VEGF₁₆₅-CM; P < 0.05). Therefore VEGF₁₆₅b did not stimulate HUVEC proliferation and, moreover, significantly inhibited VEGF₁₆₅-stimulated proliferation. Furthermore, when endothelial cells were incubated in CM containing increasing concentrations of VEGF₁₆₅b, there was a dose-dependent inhibition of [³H]thymidine incorporation stimulated by commercially available VEGF₁₆₅ (Fig. 7,b), with a molar ratio IC₅₀ of 0.94 (i.e., equimolar inhibition). Additionally, VEGF₁₆₅b-CM did not affect fibroblast growth factor-mediated proliferation (Fig. 7 c).

To determine whether VEGF₁₆₅b could inhibit other actions of VEGF₁₆₅ associated with angiogenesis, we also measured the effect of VEGF₁₆₅b on VEGF₁₆₅-mediated migration of HUVECs (Fig. 8). We incubated HUVECs in VEGF₁₆₅-CM (33 ng/ml VEGF₁₆₅), and this resulted in a significant 24% ± 3% increase in migration of endothelial cells compared with pcDNA3-CM alone (P < 0.01, ANOVA). VEGF₁₆₅b-CM containing 33 ng/ml VEGF₁₆₅b did not stimulate migration (−3 ± 2.6% compared with pcDNA3-CM, not significant). Furthermore, there was no increase in migration when cells were incubated in a combination of both VEGF₁₆₅b-CM and VEGF₁₆₅-CM containing 33 ng/ml of each isoform (9.9 ± 5.8% compared with pcDNA3-CM). Therefore, VEGF₁₆₅b did not stimulate migration and again significantly inhibited VEGF₁₆₅-stimulated migration (P < 0.001, ANOVA).

We measured the effects of VEGF₁₆₅b on vasodilatation to determine whether VEGF₁₆₅b could inhibit the effects of VEGF₁₆₅ on intact vessels. Luminal perfusion of isolated pressurized rat mesenteric arteries in vitro with dialyzed pcDNA3-CM did not affect vessel diameter (Fig. 9,a). Perfusion of the same arteries with dialyzed VEGF₁₆₅b-CM (40 ng/ml) did not affect the diameter of the arteries either (Fig. 9,a). Perfusion with dialyzed CM containing 20 ng/ml VEGF₁₆₅ resulted in significant vasodilatation (Fig. 9,b), but this vasodilatation was abolished when perfused with VEGF₁₆₅b-CM and VEGF₁₆₅ (40 ng/ml VEGF₁₆₅b; 20 ng/ml VEGF₁₆₅). Therefore, VEGF₁₆₅b does not stimulate vasodilatation and inhibits VEGF₁₆₅-mediated vasodilatation (Fig. 9 b).

VEGF expression has previously been thought to be required for tumor growth. Although this is still evidently true, the data presented here suggest that regulation of splicing of VEGF may also be important in tumor development. We have discovered a novel endogenously produced VEGF mRNA from human kidney that appears to be down-regulated in tumors. It appears somewhat surprising that this isoform has not been described previously. However, the only method that would successfully distinguish between exon 9-containing isoforms and exon 8-containing isoforms is that of RT-PCR using primers that flank the exon 9-containing region and discrimination by high-resolution gel electrophoresis. The vast majority of papers examining VEGF expression have used primers that are complementary either to the regions 5′ to exon 9 or that produce products that would not distinguish VEGF₁₆₅b from other isoforms such as VEGF₁₄₈ or VEGF₁₄₅. Detection of mRNA using Northern blotting, RNase protection assay, or in situ hybridization would not result in a discrimination between VEGF₁₆₅b and other isoforms of VEGF, because all other isoforms contain the exon 9 sequence in their 3′ UTR. A search of the nucleotide database, however, provides interesting information on the conservation of this region. The entire 3′ UTR from the end of exon 8 is 100% conserved between human and macaque. In other mammalian species, there is relatively good conservation of the supposed 3′ UTR terminal to exon 8 (Fig. 10,a). In fact, in the cow the mRNA has >95% identity in the 66 bases 3′ to the stop codon of exon 8, and exon 9 is >90% identical. However, this identity breaks down immediately after the exon 9 stop codon (Fig. 10,a). There is significantly less identity in the 22 bases after this stop codon (53%) than that in exon 9 (91%). This pattern is also evident in mouse, where the exon 9-containing sequence is 86% identical to human but only 23% identical in the 22 bp immediately after the exon 9 stop codon (Fig. 10,a). Interestingly, the mouse sequence predicts an exon 9 of 7 amino acids of the sequence PLTGKTD, compared with SLTRKD in the human and macaque and RLTRKD in the cow. Therefore, four of six amino acids are conserved [XLTXK(X)D], and this appears to have been brought about by a double mutation, an adenosine insertion in the human at nucleotide 10 (mouse) and a cytosine to thymidine mutation at nucleotide 19 (mouse) that rescues the stop codon (Fig. 10,a). This is indirect evidence for functional relevance of the splice site. Interestingly, it is only conserved in mammals, not in birds or fish (Fig. 10 b).

There are a number of particularly interesting features of this new splice variant. The splice variant contains normal exons 1–5 and 7. Exon 1 encodes for the signal sequence; therefore, it is likely that this isoform is secreted (24). This was confirmed by our Western blot showing expression in CM. Exons 3 and 4 contain the dimerization domain (25) and receptor binding domains (26). These appear to be intact in this isoform, indicating that it is capable of homodimerization and potentially heterodimerization with other isoforms. More importantly perhaps, it should be able to bind to its receptors effectively. However, Keyt et al. (16), have demonstrated that the COOH-terminal of the VEGF protein confers mitogenic activity. Digestion of VEGF₁₂₁ with plasmin results in a 110-amino acid molecule that lacks exon 8 but contains exons 1–4 and most of exon 5. This results in a 110-amino acid VEGF that is less potent at increasing human umbilical vein endothelial cell number, both as a homodimer and as a heterodimer. However, the potency of VEGF in culture is not affected by the loss of exon 8. VEGF₁₂₁, which has exons 1–5 and 8, has the same effect as VEGF₁₁₀, indicating that it is not the loss of exon 8 but rather those amino acids coded for by exon 7 that specify the ability to increase cell number (in itself a measure of the balance between survival and proliferation). VEGF₁₆₅b, which contains exon 7, should therefore have the same potency as VEGF₁₆₅.

VEGF₁₆₅b has an alternate exon instead of exon 8, however. Exon 9 encodes for 6 amino acids and therefore results in a protein of the same length as VEGF₁₆₅, although the six amino acids are very different. Exon 8 encodes for Cys-Asp-Lys-Pro-Arg-Arg (CDKPRR). The cysteine forms a disulfide bond with Cys-146 in exon 7 (27). This results in the COOH terminus of VEGF₁₆₅ being held close to the VEGF-R2 receptor binding domain in exon 4 in the three-dimensional structure of VEGF. In addition, the proline inserts a kink in the amino acid backbone of the molecule. Finally, the two terminal amino acids are highly positively charged arginine residues. Exon 9, on the other hand, codes for Ser-Leu-Thr-Arg-Lys-Asp (SLTRKD) and has lost the cysteine residue; therefore, the COOH terminus may not be held close to the receptor binding site. In addition, there is no proline kink, and there is no net charge on the terminal two amino acids. We have shown that this novel exon confers inhibitory activity upon the isoform. There are two obvious mechanisms by which this may occur. Either VEGF₁₆₅ is able to bind to the VEGF receptors but not activate them, or it heterodimerizes with VEGF₁₆₅ and prevents VEGF₁₆₅-mediated stimulation of the receptors. VEGF₁₆₅b contains all of the elements required for efficient dimerization: Cys-51 and Cys-60 in exon 3 (25); and receptor binding, for VEGF-R1 Asp63, Glu64, Glu67 in exon 3; for VEGFR2 Arg82, Lys84, His 86 in exon 4 (26); and for neuropilin-1 Cys-136 to Cys-158 in exon 7 (28). This isoform substitutes exon 9 for exon 8 and therefore has no Cys-160, which normally binds to Cys-146 in exon 7 and will therefore be likely to affect folding and tertiary structure of the VEGF molecule.

It is particularly interesting that this isoform has been found in the kidney. The kidney, and in particular the glomerulus, has been shown to produce high levels of VEGF. This has always been assumed to be VEGF₁₆₅, because it is detected by antibodies to VEGF₁₆₅ (29), in situ hybridization using probes to VEGF₁₆₅ (30), PCR using primers specific to exon 8-containing isoforms (7), and so on. However, in all of these cases, the detection techniques used would not easily distinguish between VEGF₁₆₅ and VEGF₁₆₅b. The only reason we detected VEGF₁₆₅b was that we were attempting to examine VEGF₁₄₈ expression in this tissue and came across it serendipitously. Antibodies and in situ hybridization probes to VEGF₁₆₅ may have detected VEGF₁₆₅b inadvertently. It is interesting to note that there has been an outstanding question concerning the role of VEGF in the kidney, i.e., that despite high levels of VEGF produced by podocytes in the glomerulus, glomerular endothelial cells (which express VEGF receptors) are no more proliferative than other endothelial cells. Endothelial cell turnover in the glomerulus is low, as in other parts of the body, and there is no overt angiogenesis occurring. VEGF₁₆₅b could explain this apparent paradox, high VEGF expression but low angiogenesis. Kidney tumors must grow in an environment in which VEGF is highly expressed under normal conditions, and yet angiogenesis is prevented. Therefore, tumors need to overcome some endogenous anti-angiogenic process to grow. If VEGF₁₆₅b were to be inhibitory, then the down-regulation of VEGF₁₆₅b by tumors that we have described here would be necessary for tumors to switch angiogenesis on in this tissue. Furthermore, the high production of VEGF by the glomerulus may actually be VEGF₁₆₅b rather than VEGF₁₆₅, and therefore this would explain why angiogenesis is not seen. The obvious next questions therefore are what is the relative abundance of VEGF₁₆₅b and VEGF₁₆₅ and what does VEGF₁₆₅b do if it is not angiogenic? These questions have yet to be answered.

Finally, the fact that we detected exon 9-containing isoforms in a number of other tissues shows that these isoforms may be widely distributed in the body. Of particular note was the difference in expression between aorta and inferior vena cava and between pituitary and hypothalamus, because each of these pairs of tissues came from the same patients. It is also of particular interest because the pituitary gland has a large number of fenestrated capillaries (31), whereas the hypothalamus is considered to have a vasculature that forms a normal blood brain barrier. The glomerular capillaries are also fenestrated, and it may be that VEGF₁₆₅b is necessary for formation of fenestrated capillaries, as has been shown for VEGF (32), but without inducing angiogenesis.

There are now >20 angiogenesis inhibitors undergoing clinical trials (2), many with encouraging results. However, most of these are potentially immunogenic (monoclonal anti-VEGF antibody for example) or are synthetic agents with potentially unpredictable side effects (e.g., thalidomide). Should functional studies confirm VEGF₁₆₅b and its counterparts as inhibitory, their potential use as angiogenesis inhibitors would have the theoretical benefit of using molecules that appear to be expressed endogenously. We suggest therefore that the functional capacity of this new VEGF isoform requires further investigation.