Introduction
Rheumatoid arthritis (RA) has a prevalence of about 1% in most parts of the world. While targeting TNFα using biological inhibitors has been an undoubted success, efficacy does not usually approach remission. Moreover, increasing usage of anti-TNFα biological agents in RA is associated with an augmented risk of infections, including tuberculosis [
1‐
5]. As a consequence, initiatives to develop alternative targets in RA are desirable, especially for use in combination with TNFα inhibitors.
Cell surface receptors such as receptor tyrosine kinases (RTKs) mediate ligand-induced signal transduction from the extracellular to the intracellular environment. Dysregulation of RTK signaling is implicated in the pathogenesis of many human diseases, including cancer and autoimmune diseases [
6,
7]. The discovery that soluble forms of receptors can abrogate receptor–ligand interaction has fueled substantial interest in their potential application as biotherapeutics. Etanercept, a molecularly engineered fusion protein composed of the extracellular domain of TNF receptor type II, is an example of a clinically effective soluble receptor-based therapeutic, with potent activity in RA [
8].
Soluble receptors are known to occur naturally
in vivo [
9]. Two major mechanisms involved in the formation of naturally occurring soluble receptors are proteolytic cleavage of membrane receptors and alternative pre-mRNA splicing. The latter is a process in which multiple proteins are created from a single pre-mRNA [
10‐
13]. Bioinformatics analyses predict that the majority of human genes undergo alternative splicing, suggesting that alternative splicing is a significant component in generating diversity of function in the human genome [
11]. The protein products of alternative splicing may serve as homeostatic regulators in physiology and disease [
14‐
16]. This is illustrated by the splice variant of vascular endothelial growth factor receptor (VEGFR) type 1 (sVEGFR1 or sFlt-1). Vascular endothelial growth factor (VEGF) plays a pivotal role in regulating angiogenesis, and binds sFlt-1
in vivo. Suppression of endogenous sFlt-1 was found to abolish corneal avascularity in mice [
17]. Conversely, sFlt-1 has been shown to modulate disease in other
in vivo models, including animal models of RA [
18‐
22].
To determine the frequency of functional soluble splice forms of cell surface receptors, we have developed a high-throughput method for gene scanning, cloning, and characterization that identified functional alternative splice variants (ASV). The present work describes the RT-PCR selection and molecular cloning of 60 novel soluble receptors as splice variants of 21 RTKs and other cell surface receptor genes, including VEGF and TNF receptors. These cell surface receptor-derived ASV differ from transmembrane proteins, or shed receptors, by the deletion or addition of unique amino acids as a result of alternative splicing events, including exon extensions and deletions. The novel ASV that we identified included splice variants of receptors for VEGF (VEGFR1, VEGFR2 and VEGFR3) and for angiopoietin-1 (Ang-1) receptor Tie1 (tyrosine kinase with immunoglobulin and epidermal growth factor homology domains 1), as well as for platelet-derived growth factor receptor beta (PDGFRβ) and fibroblast growth factor receptors (FGFRs).
We selected 10 ASV for further analysis, chosen on the basis of their potential effects on angiogenesis, which represent an attractive target for therapy in RA [
23‐
28]. We confirmed that ASV derived from cell surface receptors retained their ligand binding ability and were transcribed in human normal and malignant tissues. Furthermore, using adenoviruses expressing secreted ASV, we demonstrated that these ASV exhibit differential effects in a murine model of RA – namely, collagen-induced arthritis (CIA), which is in widespread use as a tool for developing new therapeutics. Work in the acute CIA model formed the basis for the widespread clinical use of TNFα inhibitors for treatment of RA [
29‐
32]. Moreover, we and other workers have shown that inhibition of angiogenesis ameliorates disease [
18,
20,
33‐
38]. We observed that ASV corresponding to VEGFR1, and to a lesser extent VEGFR2, reduced arthritis severity, in agreement with our earlier findings using sFlt-1 [
18,
20]. We also observed for the first time that ASV corresponding to Tie1 significantly reduced arthritis severity and joint destruction. While expression of Ang-1 [
39,
40] and of Tie receptors [
41‐
43] has been reported in RA, this is the first demonstration that Tie1 is effective in an
in vivo model of arthritis. We also observed a modest effect of FGFR1 ASV in acute CIA.
These data establish that ASV derived from receptors that play key roles in angiogenesis – VEGFR1 and, for the first time, Tie1 – can reduce arthritis severity. More broadly, ASV are a source of novel proteins with therapeutic potential in diseases in which angiogenesis and cellular hyperplasia play a central role, such as RA.
Materials and methods
Materials
Human umbilical vein endothelial cells (HUVEC) and endothelial cell medium-2 were obtained from Cambrex (East Rutherford, NJ, USA). Tie1-751 was 125I-custom-labeled by GE-Amersham (Piscataway, NJ, USA). Anti-human Tie1 (C18) and Tie2 (C-20) rabbit polyclonal antibodies specific to the C-terminal receptor domains were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Mouse penta-His antibody was obtained from Qiagen (Valencia, CA, USA). Anti-Myc mouse monoclonal antibody (9E10) was obtained from Roche Diagnostics (Indianapolis, IN, USA). Antibodies detecting extracellular domains of soluble receptors, human VEGFR1/Fc and VEGFR3/Fc chimeras, human VEGF-C, VEGF-D and VEGF165, and anti-human VEGF-D polyclonal antibody were obtained from R&D Systems (Minneapolis, MN, USA).
RT-PCR cloning of novel alternative splice variants and generation of alternative splice variant adenoviruses
mRNAs that represent major human tissue types from healthy or diseased tissues and from cell lines were purchased from Clontech (Mountain View, CA, USA) and from Strategene (La Jolla, CA, USA), and were pooled. Synthesis of the first-strand cDNA was performed using STRATASCRIPT reverse transcriptase (Stratagene) following the manufacturer's instructions. For PCR amplification, gene-specific PCR primers were selected. The forward primers flanked the start codon. The reverse primers were selected from the transmembrane region of the receptors. PCR conditions were 35 cycles of 95°C for 45 seconds, 60°C for 50 seconds, and 72°C for 5 minutes. The reaction was terminated with an elongation step of 72°C for 10 minutes.
PCR products were electrophoresed on 1% agarose gel, and were stained with Gelstar (BioWhittaker, Walkersville, MD, USA). The DNA bands were extracted with the QiaQuick® gel extraction kit (Qiagen), ligated into the pDrive UA-cloning vector (Qiagen), and transformed into Escherichia coli. Recombinant plasmids were selected on bacterial agar plates containing 100 μg/ml carbenicillin. For each transfection, 200 to 1,000 colonies were randomly picked and their cDNA insert sizes were determined by PCR with M13 forward vector and reverse vector primers. Representative clones from PCR products with distinguishable molecular masses as visualized by fluorescence imaging (Alpha Innotech, San Leandro, CA, USA) were completely sequenced.
For the bioinformatics analyses, computational analysis of alternative splicing was performed by alignment of each cDNA sequence to its respective genomic sequence using SIM4 (software for analysis of splice variants; Pennsylvania State University, Centre County, Pennsylvania, USA). Only transcripts with canonical (for example, GT–AG) donor–acceptor splicing sites were considered for further analysis.
The replication-deficient adenoviral expression system ViraPower was used for subcloning and expression of the ASV proteins following the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA). Recombinant ASV-expressing adenoviruses were produced and amplified in HEK293A cells (Invitrogen), purified through a double-cesium chloride centrifugation procedure, and titrated by measuring the plaque-forming units or the infectious particle units in HEK293 cells. The Adv-Fc control virus, expressing a murine IgG
2a Fc fragment, has been previously described [
44]. Adv-LacZ virus was purchased from Welgen (Worcester, MA, USA).
Alternative splice variant mRNA expression
Expression of ASV mRNA was analyzed using RT-PCR and quantitative RT-PCR. Human normal RNA and tumor RNA (Total RNA Master Panel II) was purchased from Clontech and was DNase treated. First-strand cDNA was synthesized using the ABI High Fidelity Kit (Applied Biosystems, Foster City, CA, USA). For PCR amplification, the primers were designed using Oligo 6 (Molecular Biology Insights, Inc., Cascade, CO, USA). The condition for PCR amplification of FGFR4 and FGFR4-ASV was 30 cycles of 95°C for 45 seconds, 60°C for 50 seconds, and 72°C for 1 minute. The reaction was terminated with an elongation step of 72°C for 10 minutes.
For quantitative RT-PCR, gene-specific primers and probes were designed and assayed for specificity and efficiency using a human universal RNA sample. Quantitative RT-PCR was performed using an ABI 7900 HT sequence detection system (Applied Biosystems, Foster City, CA, USA) and TaqMan
® chemistries. cDNA was amplified in triplicate wells for both the normal and variant gene on the same plate. Cycle threshold values were determined and the average cycle threshold values were calculated and analyzed using The Institute for Genomic Research, TIGR Multiexperiment Viewer hierarchical clustering module [
45].
Protein expression and secretion
Splice variant cDNAs were subcloned into pcDNA3.1 (Invitrogen) with a Myc-His tag fused at the C-terminus of the proteins. To facilitate secretion, the native signal sequences of ASV derived from Met, FGFR1, VEGFR1, and RAGE were replaced by the tissue plasminogen activator signal/pro sequence (GenBank accession number NM_000930) by PCR cloning. All constructs were sequence verified, and were transiently expressed in HEK293 cells using LipofectAmine 2000 following the manufacturer's instruction (Invitrogen). Cell culture supernatants were collected 48 hours after transfection. To analyze expression of the recombinant proteins, equal amounts (20 μl) of supernatants were separated on SDS-PAGE gels. The separated proteins were transferred to nitrocellulose membranes, and were probed with anti-Myc antibody.
Purification of recombinant Tie1-751
Tie1-751 was subcloned into pcDNA3.1 as described above with a Myc-His tag fused at the C-terminus of the proteins (Tie1-751(6His)). To construct Tie1-751-Fc, the Fc fragment of human IgG1 (from Pro100 to Lys330) was PCR amplified and fused inframe to the 3' end of Tie1-751 in the pcDNA 3.1 vector via restriction digestion using the XhoI-AgeI site. Tie1-751(6His) and Tie1-751-Fc were transiently expressed in HEK 293 cells. Conditioned media were collected 72 hours later. Tie1-751(6His) was purified using a Ni-Sepharose 6 Fast Flow column (GE-Amersham, Piscataway, NJ, USA) and Tie1-751-Fc was purified using a Protein-A Sepharose column (GE-Amersham), following the manufacturer's instructions. Purity of the recombinant proteins was >95% as determined by SDS-PAGE and Coomassie Blue staining.
Ligand binding
To determine whether the ASV bound their cognate ligands, 96-well assay plates were coated with VEGF-A, VEGF-C, platelet-derived growth factor (PDGF)-AB, hepatocyte growth factor, colony-stimulating factor (CSF), and Ang-1, respectively, at 4 μg/ml in PBS. The immobilized ligand-coated plates were used for binding of matched ASV in the same order, as follows: VEGFR1-541, VEGFR2-712, PDGFRβ-336, Met-877, CSF1R-306, and Tie1-751. In the case of VEGFR1-541, VEGFR2-712, PDGFRβ-336, Met-877, and CSF1R-306, supernatants from the ASV-expressing HEK293 cells were used for binding assays. The purified Tie1-751(6His) was used for Ang-1 binding. Binding was performed for 1.5 hours at room temperature followed by three rapid rinses in PBS/0.05% Tween-20. Bound ASV were detected using biotin-labeled, extracellular domain-specific antibodies.
Binding of Tie1-751 to human umbilical vein endothelial cells
For cell surface binding of 125I-Tie1-751(6His), HUVEC were seeded into a 96-well plate at 1.4 × 104 cell/well in endothelial growth medium-2. Next day, medium was replaced with an ice-cold binding buffer (Hanks' balanced salt solution supplemented with 20 mM Hepes and 0.25% bovine serum albumin, pH 7.5). 125I-Tie1-751 was added to the binding buffer in the presence or absence of unlabeled Tie1-751. Binding was performed at 4°C for 1 hour followed by four washes with ice-cold PBS/0.05% Tween-20. A scintillation cocktail OptiPhase 'SuperMix' (PerkinElmer, Waltham, MA, USA) was added to each well, and the plates were read by Microbeta Trilux (PerkinElmer).
For direct binding of Tie1-751 to transmembrane Tie1 and Tie2, HUVEC were seeded into a six-well plate at 0.5 × 106/well in endothelial growth medium-2. Next day, binding was carried out at 4°C for 1 hour in an ice-cold binding buffer (as above) containing 1 μM purified Tie1-751(6His). At the end of the binding, cells were treated with or without the membrane-impermeable chemical amine-reactive cross-linking agent DTSSP (3,3'-dithiobis [sulfosuccinimidylpropionate] (Pierce Biotechnology Inc., Rockford, IL, USA) at 1 mM for 30 minutes. This treatment was followed by inactivation of 3,3'-dithiobis(sulfosuccinimidylpropionate) with 20 mM Tris buffer, pH 7.5, for 15 minutes. Cells were subsequently lysed and immunoprecipitated using a C-terminal-specific anti-Tie1 or anti-Tie2 antibody. The immunoprecipitated proteins were analyzed by western blotting using anti-His antibody that recognizes the His-tagged Tie1-751.
Evaluation of the therapeutic potential of alternative splice variants in a mouse model of arthritis
Ten-week-old DBA/1-Ola/Hsd mice (H-2
q haplotype; Harlan Laboratories UK Limited, Bicester, Oxon, UK) were immunized with purified bovine type II collagen prepared inhouse, and were emulsified with Freund's complete adjuvant, containing paraffin oil, and lyophilized
Mycobacterium tuberculosis H37 Ra (Difco Becton Dickinson, Oxford, UK) [
46]. Onset of arthritic disease was around 2 to 3 weeks later. ASV adenoviruses were administered intravenously (10
7 plaque-forming units/0.1 ml per mouse) via tail vein injection to mice on day 1 of arthritis.
All limbs were assessed daily and scored as follows: 1 = slight edema or erythema; 1.5 = edema and erythema involving at least some digits; 2 = frank edema/erythema involving the entire paw; and 2.5 = pronounced edema and erythema leading to incapacitated mobility [
37,
38]. A spring-loaded caliper (least detectable measure = 0.1 mm; Rohm GB Limited, Kingston-Upon-Thames, UK) was employed to measure the hind-paw thickness (mm) daily, which was expressed as the degree of paw swelling from day 1 of arthritis (Δ
mm).
All murine work procedures had the approval of the local ethical review process committee, which followed the Helsinki Declaration Principles, and were carried out under Project Licence 70/5446.
For pharmacokinetic analysis, mice received tail vein injection of 1 × 109 plaque-forming units of Adv-Tie1-751(6His). Sera were taken after injection at the indicated times and were analyzed by SDS-PAGE followed by western blotting with anti-Tie1 antibody. Signals exposed onto an X-ray film in a visually estimated linear range were scanned and quantitated using Typhoon Trio instrument (GE-Amersham) and were compared with a known concentration of purified Tie1-751(6His).
Histological evaluation of joint architecture
At the end of the 10-day period of monitoring, the hind feet of the mice were fixed in 10% buffered formalin solution, decalcified (Rapid-Cal™; BBC Biochemical, Dallas, TX, USA), embedded in paraffin wax positioned laterally and sagittally sectioned. Serial sections of 5 to 6 μm thickness were obtained, dewaxed and stained with H & E or toluidine blue.
The stained sections were scored for changes to joint architecture by an observer blinded to the study groups. Each section was screened for changes to the joint architecture, and every joint was scored as follows: normal; mild (minimal synovitis, some cartilage loss, shrinkage in the size of cartilage chondrocytes with denucleation, and bone erosions limited to discrete foci); moderate (more extensive synovial hyperplasia, destruction of large segments of the cartilage and considerable bone erosions caused by an invasive pannus front); and severe (complete destruction of the joint architecture).
Statistical analysis
P values were determined using a two-tailed t test assuming unequal variances. Data on the progression of arthritic disease were analyzed using two-way analysis of variance. Histology data were analyzed by the chi-square test for trend.
Discussion
The proliferative and invasive nature of RA synovium has frequently led to comparisons with tumor development, and therefore the usefulness of VEGF blockade for treatment of certain cancers might be extrapolated to RA. Heterologous CIA in mice shares many features with RA, and has been widely used to study mechanisms involved in the arthritic process and to identify new strategies for RA treatment, such as TNFα inhibitors.
VEGF inhibition has been the focus of considerable clinically oriented research, and angiogenesis blockade has been shown to be effective in different
in vivo models of arthritis, including CIA [
18,
20,
36,
49,
50]. VEGF inhibition
in vivo, however, is associated with side effects, such as impaired wound healing, hemorrhage, and gastrointestinal perforation. This is not surprising, given the heterozygous lethal phenotype of VEGF knockout mice [
51], which suggests a strategic role for this molecule. Other positive regulators of angiogenesis expressed in RA include hepatocyte growth factor and PDGF [
52,
53]. To date, however, there have been no concerted efforts to compare a range of different antiangiogenic approaches side by side in a single study.
Bioinformatics surveys [
11] and exon profiling [
13,
54] reveal that the majority of pre-mRNAs are alternatively spliced. As such, use of these soluble receptor variants might prove invaluable in designing new therapeutic strategies. We report here that, using an efficient approach, we cloned 60 novel ASV of 21 genes encoding RTKs and other cell surface receptors. The discovery of so many novel splice variants from a small group of well-characterized drug target genes is consistent with reports suggesting that alternative splicing is one of the most significant components generating protein and functional diversity in the human genome [
13,
54,
55].
In vivo, soluble receptors are generated by both alternative pre-mRNA splicing and proteolytic cleavage (shedding) of membrane-anchored receptors, resulting in truncated molecules lacking a transmembrane domain and an intracellular segment. Soluble receptors may retain their ability to bind ligands and function as ligand antagonists [
9]; for example, soluble TNF receptors [
8] and soluble VEGFR1 [
56]. Soluble receptors are often generated through rational engineering. A major difference between splice variant-derived soluble receptors and engineered soluble receptors is that the former contains novel amino acids and domain structures typically derived from intron fusion. These alterations may subsequently alter the functionality of the ASV as compared with the engineered or metalloprotease-generated soluble receptors. An example of altered function via alternative splicing is VEGF
165b, an antiangiogenic factor derived from the alternative splicing of VEGF pre-mRNA [
57]. VEGF
165b antagonizes the angiogenic effect of VEGF
165, which is also encoded by the VEGF gene. Further studies are required to elucidate the endogenous expression and function of the ASV described in this report.
Inhibiting angiogenesis is a promising strategy for treatment of neovascularization-related diseases [
58], including RA [
26]. Prior to anti-TNF therapeutics, 50% of RA patients become moderately disabled within 2 years and become severely disabled within 10 years of disease onset. The increasing use of anti-TNFα biological agents in RA is a major step forward, but its use is restricted by an associated risk of infection, including tuberculosis [
1]. Most importantly, efficacy in longstanding treatment does not usually result in remission. As a consequence, initiatives to develop alternative treatments that control disease progression in RA are desirable.
A well-documented feature of RA is an alteration in the density of synovial blood vessels. Several angiogenic factors are expressed in RA, including VEGF, PDGF, fibroblast growth factor 1, and fibroblast growth factor 2, as well as Ang-1. Angiogenesis is a multistep process, however, and – while VEGF is important – other proangiogenic factors are also expressed in RA and CIA. The contribution of other proangiogenic factors to arthritic disease progression has not been well defined or compared directly within the same disease model. In the present study, 10 RTK-derived ASV were screened side by side in the high-throughput CIA model, using replication-incompetent adenoviruses as a delivery and in vivo expression system. This method allows for screening many candidate biologics quickly in a relevant disease model, without first expressing and purifying the target molecules, and will select for proteins that are significantly expressed and are bioactive across species barriers. Some candidate proteins may give false negative results because of issues related to expression and stability in vivo, a species barrier, or a lack of activity in the particular disease model.
In vivo screening of the ASV demonstrated clear differential effects. Among them, ASV derived from VEGFR1 and Tie1 were found to be the most potent. The effect of VEGFR1-541 ASV confirms our own previous data and that of others, demonstrating the effectiveness of VEGFR1 blockade in models of arthritis [
18,
20,
33,
50]. In contrast, blockade of VEGFR2 in models of arthritis has in general not been effective [
33,
50]. The effect of VEGFR2-712 ASV was modest in our study, with inhibition of clinical score but not of paw swelling or histological change. As the ultimate benefit of a potential therapeutic in RA would be joint protection and reduced edema, the fact that VEGFR2-712 ASV does not affect either paw swelling or joint inflammation/destruction supports the view that VEGFR2 blockade is not likely to be beneficial in RA.
Expression of both Ang-1 [
39,
40] and angiopoietin receptors Tie1 and Tie2 [
41‐
43] in RA synovial tissue has been described. Ang-1 is chemotactic and weakly mitogenic for HUVEC [
59,
60], promotes formation of endothelial sprouts [
61], and has been proposed to act in concert with VEGF to promote vascular network maturation [
62,
63]. Furthermore, Ang-1 was found to be a survival factor for endothelial cells, protecting HUVEC from apoptosis induced by serum withdrawal [
64]. Angiopoietin signaling was until recently considered to be mediated via Tie2. The embryonic lethality of Tie1 knockout mice, however, suggested that Tie1 signaling is important in vascular network formation. It is now thought that Tie1 may modulate signaling through Tie2 [
65‐
67]. Marron and colleagues reported that activation of Tie1 ectodomain cleavage increased activation of Tie2, which could potentially control signaling via Tie2 [
68].
The novel activity of Tie1-751 in the CIA model [
35,
69] motivated us to further examine its mechanism of action. Our results demonstrate that Tie1-751 directly binds to Tie1 and Tie2 on the surface of endothelial cells. Binding of Tie1-Fc to transmembrane Tie1 and the interaction of transmembrane Tie1 and Tie2 at the cell surface have recently been reported [
66]. The mechanism of binding Tie1-751 to Tie1 and to Tie2, however, is currently unknown. Our initial characterization also revealed that Tie1-751 inhibits Ang-1-induced Tie1 phosphorylation and the prosurvival effect of Ang-1 on HUVEC (data not shown). These results suggest that Tie1-751 may inhibit activation of the angiopoietin–Tie system by both sequestering ligand and forming nonsignaling heterodimers with cell surface receptors. It is possible that the C-terminal intron-encoded domain of Tie1-751 expands its functionality. Blocking Tie2 has been reported effective in CIA, but no such data are available for Tie1 inhibition [
70].
Further work is needed to confirm the function of novel domains generated by alternative splicing. The differential effects of the 10 ASV in arthritis in vivo, however, suggest that selected ASV may have potential therapeutic application in RA and in other angiogenesis-dependent conditions.
Acknowledgements
The authors sincerely thank Scott Patton for editing this manuscript, and the RBLX research team for support, discussion, and critical reading of this manuscript. They are grateful to the staff of the Biological Services Unit (Kennedy Institute of Rheumatology, Imperial College, London, UK) for help in the care and maintenance of the laboratory mice used in our studies; to the Histopathology Department, Charing Cross Hospital, London – particularly David Essex, David Peston, and Ann Sandison – for help in the sectioning and staining of mice hind feet specimens; and to Kerri Reilly and Yvonne Raatz for advice with the molecular biology studies. MF, EMP, and PFS would like to thank the Arthritis Research Campaign of Great Britain, which provided support for this work.
Competing interests
PJ, JZ, IN, BJ, and HMS are employees of Receptor BioLogix, Inc. and hold stocks in the company, and declare competing financial interests. MF and EMP have acted as consultants for Receptor BioLogix, Inc. The other authors declare that they have no competing interests. Receptor BioLogix, Inc. holds the patents related to the content of the manuscript.
Authors' contributions
HMS designed the study. PJ assisted in the study design, oversaw the project running and data analysis, and drafted the manuscript. PJ, JZ, IN, and BJ performed the alternative splice variant cloning, sequence analysis, protein expression and purification, and ligand/receptor binding assays. SP performed and analyzed the quantitative PCR experiment. EMP and MF assisted in the study design and coordination, and oversaw the data analysis and drafting of the manuscript. PFS and DC designed and carried out the in vivo arthritis studies. All authors read and approved the final manuscript.