Results

Cloning of novel alternative splice variants coding for secreted receptor isoforms
To identify novel splice variants from cell surface receptor genes, we performed RT-PCR using a complex mRNA pool representing major human tissue types and tumors. We intended to identify novel splice patterns that lead to the formation of secreted receptor isoforms. To do so, we selected forward PCR primers that flank the start codon and reverse primers that are located in the transmembrane regions. The amplified PCR products were separated on agarose gels and the DNA bands were extracted, purified, and individually cloned to generate gene-specific plasmid cDNA libraries. Two hundred to 1,000 random recombinant clones within each library were screened using PCR amplification to analyze the insert sizes. Clones with subtle differences in insert sizes on agarose gel electrophoresis were selected for complete DNA sequencing. Novel splice variants were identified by alignment of each cloned sequence to its respective genomic sequence in comparison with full-length transcripts of sequence databases of National Center for Biotechnology Information (NCBI) using the splice variant analysis software SIM4 [47]. Only transcripts with canonical donor–acceptor splicing sites (for example, GT–AG) were considered for further analysis, so that potential PCR artifacts were excluded. We defined a novel splice variant as an alteration in splice patterns to the existing full-length transcript sequences from available sequence databases, including Geneseq and other public databases.
A total of 60 full-length splice variants, derived from the extracellular domains of the 21 type 1 receptor genes, were confirmed to be novel – with variants from the c-Met proto-oncogene being the most diverse (Table 1). Sequences of the 60 full-length novel splice variants were deposited with GenBank (accession numbers EU826561 to EU826620; see also Additional files 1 and 2). Alignment of the cloned splice variant cDNA sequences with the corresponding genomic and known transcript sequences in available databases revealed that a total of 83 alternative splice events occurred in the 60 novel variants (Figure 1). We categorized the alternative splice events, and found that 67.5% led to intron fusion (intron sequences inserted into mature mRNA). These include novel exon insertion, exon extension, and intron retention. The remaining 32.5% of alternative splice events resulted in exon loss (a portion or whole exon was skipped). A total of 18% of the exon extensions and 50% of the exon truncations identified in this study occurred at the 5' end of the alternatively spliced exons. All of the 60 transcript variants encounter a stop codon within the extracellular regions. As a result, these variants encode soluble receptor isoforms, and were subsequently referred to as ASV.
Table 1  Cloned alternative splice variant mRNAs  Sixty novel alternative splice variants were cloned from 21 cell surface receptor genes by RT-PCR amplification followed by extensive colony screening. The number of novel alternative splice variants is presented for each receptor tested. NCBI, National Center for Biotechnology Information.
Figure 1  Splice events categorized by type. A total of 83 alternative splicing events were identified in the 21-gene array (Table 1). The identified splicing events fell into five listed types. The splice pattern of the known transcript is depicted as type 1.

Detection of alternative splice variant mRNA expression
Expression of ASV mRNA relative to their corresponding constitutively spliced transcripts was analyzed by both RT-PCR and quantitative RT-PCR. Amplification of each target sequence was performed across 29 distinct normal tissues as well as cancer tissues including two cancer cell lines. For PCR amplification of ASV, one primer was selected within the intron fusion sequence and the other from a remote exon encompassing several introns. This approach ensured that only the variant-specific mRNA transcript was amplified. An example of typical ASV mRNA expression (FGFR4) detected by RT-PCR is shown in Figure 2a.
Figure 2  Alternative splice variant mRNA expression. (a) RT-PCR detection of mRNA expression of FGFR4 (top panels) and FGFR4-ASV (bottom panels) across 20 normal tissues and nine cancers, including two cancer cell lines. The amplified RT-PCR products were separated on 1% agarose gels and visualized by ethidium bromide staining. bp, base pairs. (b) Expression profile heat map of the constitutively expressed (C) and matched splice variant (V) mRNAs. Transcripts were analyzed across 20 normal tissues and nine cancers, including two cancer cell lines. Amplification of the constitutive and splice variant sequences was performed using real-time PCR. Bar shows a color shift from green (high-level expression) to red (low-level expression), with the corresponding cycle threshold values indicated.   For a better comparison of mRNA expression and tissue distribution, quantitative RT-PCR was performed to analyze ASV and their corresponding constitutively spliced transcripts. Our results demonstrated that expression of seven alternative splice variant mRNAs (VEGFR1, VEGFR3, Met, RAGE, Tie1, FGFR1, and Kit) is present in multiple normal and tumor tissues (Figure 2b). Levels of expression varied among tissues, with the ASV derived from VEGFR1, Met, and FGFR1 being predominantly expressed in tumor tissues. In contrast, ASV derived from VEGFR3 had the most restricted expression, and were observed only in a few normal tissues and cancer cell lines. These preliminary results indicate that expression of ASV is tissue specific and occurs more frequently in tumor than normal tissues.

Ligand binding potential of recombinant alternative splice variants
Among the 60 ASV cloned, we selected 10 for initial functional testing (Table 2). The selected ASV (corresponding to ASV derived from VEGFR1, VEGFR2, VEGFR3, Tie1, Met, Kit, CSF1R, PDGFRβ, FGFR1, and RAGE) represent diverse members of gene families, possess known functional domains such as ligand binding domains, and encode novel amino acids compared with previously reported splice variant sequences.
Table 2  Alternative splice variants selected for functional testing  Ten alternative splice variants were selected for functional testing. aLengths of the alternative splice variant open-reading frames (ORF) and lengths of the wildtype receptor extracellular domains (ECD) are indicated by the numbers of amino acids. bNovel C-terminal amino acids of each alternative splice variant are shown. *Stop codon.  Efficient expression and secretion of the selected 10 recombinant ASV (VEGFR1, VEGFR2, VEGFR3, Tie1, Met, Kit, CSF1R, PDGFRβ, FGFR1, and RAGE) from HEK293 cells was confirmed by western blot analysis of the cell culture supernatants, using anti-Myc antibody to detected the Myc-tagged ASV (Figure 3a). Furthermore, we observed ligand binding by ASV proteins derived from VEGFR1, VEGFR2, PDGFRβ, Met, and CSF1R – which bound to VEGF-A, VEGF-C, PDGF, hepatocyte growth factor, and CSF, respectively (Figure 3b). For evaluation of Tie1-751, purified recombinant protein was used for binding to Ang-1, and a dissociation constant (Kd) of approximately 89nM was measured (Figure 3b).
Figure 3  Expression and ligand binding of recombinant alternative splice variants. (a) HEK293 cells were transiently transfected with the indicated cDNA constructs. Conditioned media of HEK293 cells were collected after 48 hours, separated on SDS-PAGE gels and probed with an anti-Myc antibody to detect the Myc-tagged alternative splice variants (ASV). Molecular weights (kDa) are indicated. (b) For VEGFR1-541, VEGFR2-712, PDGFRβ-336, Met-877 and CSF1R-306, conditioned media from untransfected (Control, dashed lines) or ASV-transfected (Specific, solid lines) HEK293 cells were applied to plates precoated with the receptor-specific ligands. Unbound ASV were detected using antibodies against the extracellular domains of the receptors. Purified Tie1-751(6His) was used for Ang-1 binding, as above. Kd, dissociation constant. (c) Solution binding of VEGF-D to VEGFR3-765-Myc. Binding was carried out by combining VEGF-D with conditioned medium from either VEGFR3-765-Myc-expressing cells (lanes 1 to 3) or untransfected cells (lane 4). Subsequent immunoprecipitation was performed using anti-VEGF-D antibody and detected using anti-Myc antibody. To confirm the specificity of interaction between VEGF-D and VEGFR3-765-Myc, binding was performed in the presence of fivefold molar excess of either recombinant human VEGFR3/Fc chimera (lane 2) or soluble recombinant human VEGFR1/Fc chimera (lane 3). Molecular weights (kDa) are indicated. CM, Conditioned medium; IP, Immunprecipitation; WB, Western blot.   Not all receptor–ligand interactions could be detected by plate-based binding, which may be a consequence of steric issues associated with binding receptor or ligand to the surface of the plate. Binding of VEGF-D to VEGFR3-765, for example, was demonstrated only when the assay was performed in solution (Figure 3c). Specificity of VEGF-D binding to VEGFR3-765 was confirmed using a soluble VEGFR3/Fc chimera, which was able to compete with VEGFR3-765 binding to VEGF-D – unlike a soluble VEGFR1/Fc chimera (Figure 3c).

Tie1-751 binds to membrane Tie1 and Tie2 on human umbilical vein endothelial cells
Some soluble receptor splice variants have been shown to bind cognate cell surface receptors and to modulate response to ligand [48]. Tie1-751 comprises most of the extracellular domain of Tie1 plus 11 C-terminal intron-derived amino acids. To begin understanding the functionality of Tie1-751, we tested whether Tie1-751 binds to endothelial cells. Proliferating endothelial cells (HUVEC) were incubated with 125I-labeled Tie1-751. Our results showed that 125I-Tie1-751 specifically bound to HUVEC, with an estimated dissociation constant (Kd) of 121 nM (Figure 4a). Binding of 125I-Tie1-751 to HUVEC was competed by increasing amounts of unlabeled Tie1-751 (Figure 4b).
Figure 4  Tie1-751 interacts with Tie1 and Tie2. (a) Specific binding of 125I-Tie1-751(6His) to human umbilical vein endothelial cells (HUVEC). Nonspecific binding was determined in the presence of 100-fold excess of unlabelled Tie1-751 and was subtracted from the total binding. CPM, counts per minute; Kd, dissociation constant. (b) Binding of 125I-Tie1-751(6His) to HUVEC was competed by increasing amounts of cold Tie1-751. Data are the mean ± standard error of the mean. (c) Binding of Tie1-751(6His) to HUVEC. At the end of binding, cells were treated with or without the cross-linker 3,3'-dithiobis(sulfosuccinimidylpropionate) (DTSSP), immunoprecipitated using a C-terminal-specific anti-Tie1 (top panel) or anti-Tie2 (middle panel) antibody, and were analyzed by western blotting using anti-His antibody. To confirm equal loading, cell lysates were blotted with anti-Tie1 antibody (bottom panel). IP, Immunprecipitation; WB, Western blot.   Direct binding of Tie1-751(6His) to Tie1 and Tie2 on HUVEC was also examined. Our results demonstrated interaction of Tie1-751(6His) with the transmembrane Tie1, as well as with the transmembrane Tie2 (Figure 4c).

Evaluation of alternative splice variant activity in an in vivo model of arthritis
Since angiogenesis plays a key role in RA, we next evaluated the therapeutic potential of ASV in an extensively validated mouse model of arthritis – namely, acute CIA. On the day of disease onset, replication-incompetent alternative splice variant-expressing adenoviruses were administered as a single dose of 1 × 107 plaque-forming units. The severity of arthritis in the mice was consecutively recorded for the following 10 days.
Control adenovirus (LacZ) was without significant effect on disease severity (Table 3 and Figures 5 and 6). In contrast, treatment with either Tie1-751 (Table 3 and Figure 5) or VEGFR1-541 (Table 3 and Figure 6) alternative splice variant adenoviruses significantly reduced disease severity, as evidenced by decreased clinical scores (P < 0.001), reduced paw thickness (P < 0.001), and reduced joint inflammation and destruction (P < 0.01 and P < 0.001 for VEGFR1-541 and Tie1-751, respectively). An example of the joint histology for untreated, LacZ ASV-treated and Tie1-751 ASV-treated mice is shown in Figure 5c, with quantitative analysis of the histology depicted in Table 3.
Table 3  Effect of alternative splice variant-expressing adenoviruses on joint inflammation and destruction  Following onset of arthritis, mice were treated with the alternative splice variant adenovirus indicated. Data presented as P values of mice treated with the indicated recombinant alternative splice variant adenoviruses as compared with untreated mice, and are expressed as the P value of clinical scores, paw swelling, and histological evaluation. For clinical scores and paw swelling, data were analyzed using two-way analysis of variance versus untreated mice. For histological evaluation, H & E and toluidine blue stained sections were scored for pannus formation, synovitis, and bone and cartilage erosion. Data were analyzed using the chi-square test for trend versus untreated mice.
Figure 5  Inhibition of murine collagen-induced arthritis by Tie1-751. On the day of arthritis onset, mice received intravenously 1 × 107 plaque-forming units of adenoviruses expressing either LacZ (○) or Tie1-751 alternative splice variants (ASV) (●), or remained untreated (□) as indicated. (a) Clinical score was recorded daily, and data were analyzed by two-way analysis of variance versus untreated mice. LacZ, not significant (P = 0.3734); Tie1-751, P < 0.001; n = 6 per group. (b) Paw swelling was recorded using calipers daily, and data were analyzed by two-way analysis of variance versus untreated mice. LacZ, not significant (P = 0.5134); Tie1-751, P < 0.001. Data are means of n = 6. (c) Serial sections of mouse hind feet were stained with either H & E (left panels) or toluidine blue (right panels). Figure shows tibia–tarsus joint sections from untreated mice (top panels), from LacZ adenovirus-treated mice (middle panels), and from Tie1-751 ASV adenovirus-treated mice (bottom panels). Sections are shown at 40× magnification; scale bar = 20 μm. (d) Pharmacokinetics of Tie1-751 from the ASV-expressing adenovirus. Sera from untreated mice or mice treated intravenously with 1 × 109 plaque-forming units of Tie1-751 ASV adenovirus were analyzed after the indicated times by western blot, followed by scanning and quantitation using Tie1-751 standard. (e) Effect of recombinant Tie1-751-Fc protein on clinical score. Results are from mice on day 10 of arthritis. Filled bars, untreated mice; empty bars, mice treated with recombinant Tie1-751-Fc 30 mg/kg, three times weekly. Data are means of n = 6. **P < 0.01 for Tie-751-Fc treated mice versus untreated mice.
Figure 6  Differential effects of alternative splice variant-expressing adenoviruses on collagen-induced arthritis mice. On the day of arthritis onset, mice received intravenously 1 × 107 plaque-forming units of the indicated alternative splice variant (ASV)-expressing adenoviruses. Clinical scores ((a), (c), and (e))and paw thickness measured by calipers ((b), (d), and (f))were recorded daily. Data were analyzed by two-way analysis of variance versus untreated mice (Table 3). (a) and (b) Mice received adenoviruses expressing either LacZ (○), VEGFR1-541 (■), VEGFR2-712 (▲) or VEGFR3-765 (●), or remained untreated (□). Data are means of n = 5 per group. (c) and (d) Mice received adenoviruses expressing either LacZ (○), Met-877 (■), Tie1-751 (▲) or FGFR1-320 (●), or remained untreated (□). Data are means of n = 6 per group. (e) and (f) Mice received adenoviruses expressing either LacZ (○), RAGE-387 (■), PDGFRβ-366 (▲), c-Kit-413 (●), or CSF1R-306 (◆), or remained untreated (□). Data are means of n = 6 per group.   The presence of Tie1-751 in mouse sera was confirmed by western blotting (Figure 5d). The effectiveness of Tie1-751 in CIA was confirmed using recombinant Tie1-751-Fc protein (Figure 5e).
A less marked disease-modifying effect was seen with the adenovirus encoding FGFR1-320 (Table 3 and Figure 6), which reduced clinical scores and paw thickness (P < 0.01) but without achieving a statistically significant improvement of joint histological evaluation (P < 0.057). Similarly, VEGFR2-712 reduced the clinical score (P < 0.001) but failed to affect the paw thickness and the histological scores (Table 3 and Figure 6).
Treatment with ASV derived from VEGFR3, RAGE, Met, c-Kit, PDGFRβ, and CSF1R adenoviruses did not generate a significant effect on any of the disease parameters (Table 3 and Figure 6).