Drs. Shi and Ciurli contributed equally to this work.
Experimental immunity to the G1 domain of the proteoglycan versican induces spondylitis and sacroiliitis, of a kind seen in human spondylarthropathies
Article first published online: 7 OCT 2003
Copyright © 2003 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 48, Issue 10, pages 2903–2915, October 2003
How to Cite
Shi, S., Ciurli, C., Cartman, A., Pidoux, I., Poole, A. R. and Zhang, Y. (2003), Experimental immunity to the G1 domain of the proteoglycan versican induces spondylitis and sacroiliitis, of a kind seen in human spondylarthropathies. Arthritis & Rheumatism, 48: 2903–2915. doi: 10.1002/art.11270
- Issue published online: 7 OCT 2003
- Article first published online: 7 OCT 2003
- Manuscript Accepted: 13 JUN 2003
- Manuscript Received: 9 JAN 2003
- Shriners Hospitals for Children
- Riva Foundation
- Canadian Institutes of Health Research/The Arthritis Society
Experimental immunity to the G1 domain of the cartilage proteoglycan (PG) aggrecan (AG1) leads to the development of spondylitis as well as polyarthritis in BALB/c mice. The PG versican contains a structurally similar G1 domain (VG1). This study was conducted to determine whether immunity to VG1 would elicit similar pathology in these mice.
Recombinant natively folded VG1 and AG1 were prepared. BALB/c mice received either a series of 5 injections of human VG1 or AG1, or no protein. Polyarthritis was determined clinically, and spondylitis and sacroiliitis histologically. Immunohistochemistry of rat tissues was used to study the localization of versican. Enzyme-linked immunosorbent assays were employed to study humoral immunity to the recombinant proteins as well as to overlapping synthetic peptides covering all these human G1 domains and mouse homologs. Affinity-purified antibodies to human AG1 and VG1 were isolated from sera of hyperimmunized mice. T lymphocyte proliferation assays were performed using recombinant human proteins. T cell lines reactive with specific immunodominant T cell epitopes in human AG1 and VG1 were isolated. Synthetic peptides encoding sequences in these human proteins and in corresponding mouse proteins were used in these analyses. Guanidinium chloride extracts of mouse spines were also used in Western blots to study antibody cross-reactivity.
Immunity to recombinant VG1 did not result in clinical polyarthritis. There was, however, clear evidence that VG1, like AG1, could induce spondylitis in the lumbar spine and sacroiliitis. Accumulation of mononuclear cells was observed in spinal ligaments adjacent to the intervertebral disc, in the intervertebral disc, and in the sacroiliac joints, the same sites where versican is localized. In contrast to AG1-immunized mice, in which T cells reactive with human AG1 cross-reacted with mouse AG1, there was no evidence in VG1-immunized mice that T cell immunity to human VG1 was cross-reactive with a mouse synthetic peptide that contained the sequence corresponding to the single immunodominant T cell sequence recognized in human VG1. Antibodies to specific sequences in human VG1 did, however, cross-react with human AG1 and with corresponding peptide sequences in mouse versican and aggrecan and with mouse proteins containing VG1 and AG1, present in mouse spine extracts. Similarly, antibodies to human AG1 cross-reacted with human VG1 and with extracted mouse VG1 and AG1 and synthetic peptides containing mouse sequences that corresponded to the reactive human epitopes in AG1 and VG1.
These observations suggest that humoral immunity to human VG1 is involved in the induction of experimental spondylitis and sacroiliitis in BALB/c mice. This humoral immunity is cross-reactive with mouse versican and aggrecan but is not associated with polyarthritis, probably because of the lack of cross-reactive T cell immunity and the absence of detectable versican in articular cartilage limbs. Induction of polyarthritis by bovine or human aggrecan requires the involvement of immunity mediated by T lymphocytes that are cross-reactive to a mouse aggrecan epitope. Together these observations suggest that humoral immunity to versican as well as immunity to aggrecan may be of importance in the development of the spinal pathology characteristic of spondylarthropathies.
Ankylosing spondylitis (AS) is a chronic inflammatory disease of the axial skeleton and the sacroiliac (SI) joints. It involves the erosive destruction of extraarticular sites such as the entheses, aorta, and eyes (1). Peripheral joint disease develops in a minority of patients (0.5%). AS affects at least half a percent of the worldwide population (1). Relatively little is known of its causes, and its immunologic nature is poorly understood. Immunohistologic and in situ hybridization studies of the SI joints of AS patients, using samples obtained by guided needle biopsy, have revealed that the lesions contain mononuclear cell infiltrates composed of CD4+,CD8+ T lymphocytes and CD14+ macrophages (2). T cells and macrophages are associated with early active sacroiliitis (3). Autoimmunity against one or more autoantigens may be involved in the pathogenesis of AS (4). We have shown that the cartilage proteoglycan (PG) aggrecan (5, 6), and specifically the amino-terminal G1 globular domain of PG (AG1) (7), can induce spondylitis and erosive polyarthritis in BALB/c mice. Cellular immunity to PG is frequently found in AS patients (8, 9) as well as in patients with rheumatoid arthritis (10), and autoreactive CD8+ T cells directed against HLA–B27–associated self-epitopes have recently been observed in patients with AS (11).
A spondylitis-like pathology is seen in murine progressive ankylosis in mice (12), in HLA–B27–transgenic rats (13), and in BALB/c mice immunized with aggrecan (5, 6) or AG1 (7). Murine progressive ankylosis is caused by a recessive genetic defect, but there is no evidence of an immune pathogenesis of the kind seen in AS (12). B27-transgenic rats develop peripheral arthritis, enterocolitis, and genital, skin, and nail lesions (13). Animals with experimental AS also develop a peripheral arthritis (5–7, 12, 13), as is seen in a small subset of AS patients (14). Immunization with cartilage matrix link protein induces an erosive polyarthritis but, despite the considerable sequence homology between AG1 and link protein, there is no spinal arthropathy with link protein immunization (15). In view of its structural homology with the G1 domain of aggrecan and link protein and its presence in tissues that are involved in the pathology of AS, it has been suggested that the G1 domain of the PG versican (VG1) may also be a target for autoimmunity in inflammatory arthritis, especially the spondylarthropathies (16).
We show here that versican is a component of the annulus of the intervertebral disc, of spinal ligaments, of the entheses, and of the SI joints, but not of articular cartilage with the exception of the SI joints. When immunity to recombinant human VG1 is induced in BALB/c mice, there is induction of spondylitis, sacroiliitis, and enthesitis in the sites where versican is located. This pathology is similar to that seen with AG1 except that in the case of VG1 there is no clearly defined peripheral arthritis. This VG1-induced pathology is associated not with T cell, but with B cell, cross-reactive immunity between human VG1 and AG1, and mouse VG1 and AG1. These observations reveal that experimental disease with features that closely resemble those of AS can develop in response to immunity involving these PGs. Such immunity may be of importance in the pathogenesis of human spondylarthropathies.
MATERIALS AND METHODS
Inbred male and female BALB/c mice, 6–8 weeks old and weighing 16–20 gm, were obtained from The Jackson Laboratory (Bar Harbor, ME).
Preparation of the genetic constructs and recombinant donor plasmids.
To express human VG1 and AG1 as recombinant proteins, a human VG1 complementary DNA (cDNA) fragment and a human AG1 cDNA fragment were generated by reverse transcriptase–polymerase chain reaction (PCR) from a human keratinocyte RNA preparation and a human chondrocyte RNA preparation, respectively. An 18-bp sequence encoding a 6xHis tag was generated at the C-termini of VG1 and AG1 to facilitate purification with a nickel affinity column. For VG1, the genetic construct codes for the signal peptide of human versican incorporating amino acid residues 1–343 (starting at LHK and finishing at DSE) of mature human versican and the 6xHis tag. The forward primer was 5′-GGATCCAAGCCGCCTTCCAAGGCCAAGATGTTC-3′ with a Bam HI linker site (underlined) and a start codon ATG (underlined). The reverse primer was 5′-GAATTCTCAATGGTGATGGTGATGATGTTCTGAATCTATTGGATGACCAATTAC-3′ with an Eco RI linker site (underlined) followed by a translation termination codon and an 18-bp sequence for the hexaHis tag (underlined).
Human primary culture keratinocytes were purchased from Colnetics (San Diego, CA). For AG1, the genetic construct codes for the signal peptide of human aggrecan involving amino acid residues 1–412 (starting at VET and finishing at FAE) of mature human aggrecan and the 6xHis tag. The forward primer was 5′-GGATCCACTATGGCCACTTTACTCTGGGTTTTCG-3′ with a Bam HI linker site (underlined) and a start codon ATG (underlined). The reverse primer was 5′-GAATTCTCAATGGTGATGGTGATGATGCTCAGCGAAGGCAGTGGC-3′ with an Eco RI linker site (underlined) followed by a translation termination codon and an 18-bp sequence for the 6xHis tag (underlined).
Human chondrocytes were isolated from human articular cartilage at autopsy. Total RNA was extracted from keratinocytes or chondrocytes using TRIzol reagent according to the protocol provided by the manufacturer (Life Technologies, Grand Island, NY). First-strand cDNA was synthesized according to the protocol provided in the SuperScript first-strand synthesis system (Life Technologies) and used as a template. PCR was performed with a thermal cycler. We used a cycle sequence of 1 minute at 94°C, 1 minute at 55°C, and 1.5 minutes at 72°C, with the primers described above. The amplified PCR product was cloned into a pCR2.1 vector using a TA cloning kit (Invitrogen, Carlsbad, CA). After confirmation, by DNA sequencing, of the authenticity of the cDNA fragment in the pCR2.1 vector, the cDNA fragment was subcloned into the pFastBac1 vector from the Bac-to-Bac Baculovirus Expression System kit (Life Technologies) at the Bam HI and Eco RI sites.
Construction of the recombinant bacmids and baculoviruses.
The recombinant bacmids and baculoviruses were prepared according to the Bac-to-Bac Baculovirus Expression System protocol. Briefly, the recombinant pFastBac1 donor plasmid was transformed into DH10BAC-competent cells, and the recombinant bacmids were isolated from the culture of the positive DH10BAC colonies. The recombinant bacmid DNA was transfected into sf9 insect cells (Invitrogen) using CELLFECTIN reagent (Life Technologies). The supernatants containing the secreted recombinant protein and baculovirus particles were harvested at 72 hours posttransfection, for further protein characterization and virus amplification.
Expression and purification of the recombinant proteins.
The recombinant baculoviruses were amplified using sf9 cells cultured in Sf-900 II SFM medium (Life Technologies). High Five insect cells, purchased from Invitrogen, were cultured in Express Five SFM medium (Invitrogen) and infected with the recombinant virus at 10 multiplicity of infection units to produce recombinant proteins. The supernatant was collected at 3.5 days postinfection and applied to a Sephadex G-25 column equilibrated with phosphate buffered saline (PBS; pH 7.4). The protein-containing fractions were collected and applied to an Ni-NTA agarose bead column (Qiagen, Mississauga, Ontario, Canada). After washing with 40 mM imidazole in PBS (pH 7.4) containing 0.3M NaCl, the His-tagged proteins were eluted with 100 mM imidazole in PBS (pH 7.4) containing 0.3M NaCl.
Characterization of purified recombinant proteins.
To assess the purity of the recombinant proteins, 0.5 μg VG1 and 0.85 μg AG1 were analyzed by 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions, and silver stained. Protein content was determined as previously described (17).
Synthetic overlapping human and non-overlapping mouse peptides representing the sequences of VG1 and AG1 (Figure 1) were prepared at the Core Facility, Shriners Hospitals for Children. They ordinarily overlapped by 9 (AG1) or 10 (VG1) residues. Human peptides were usually 23–25 residues in length. Peptides were prepared using 9-fluorenylmethoxycarbonyl chemistry (FastMoc) on an Applied Biosystems (Foster City, CA) 431A synthesizer and purified by high-performance liquid chromatography using an Aquapore C8 column (Brownlee; Perkin-Elmer, Woodbridge, Ontario, Canada).
Hyaluronan (HA) biotinylation and HA binding activity of recombinant VG1 and AG1.
To verify whether the recombinant VG1 and AG1 expressed using the baculovirus expression system fold properly, an HA binding assay for VG1 and AG1 was used (only naturally folded protein binds HA). Biotinylated HA was prepared as previously described (18). Ninety-six–well plates (Greiner America, Lake Mary, FL) were coated with 50 μl (10 μg/ml) of recombinant human VG1, AG1, or ovalbumin (as a negative control) (Sigma, St. Louis, MO) in PBS overnight at 4°C. They were blocked with 1% bovine serum albumin (Fraction V; Sigma) in PBS for 2 hours at room temperature and washed 3 times with PBS containing 0.05% Tween 20 (PBST).
To measure HA binding activity of AG1 or VG1, 50 μl of biotinylated HA at various concentrations (0.8–800 ng/ml) was added to each well and incubated for 1 hour at room temperature. In a separate competitive binding assay, 50 μl of 1 μg/ml of biotinylated HA with increasing concentrations of unlabeled HA (0.1–1,000 μg/ml) was added to VG1- or AG1-coated wells and incubated for 1 hour at room temperature. The plates were washed 3 times with PBST, and then 50 μl of alkaline phosphatase–conjugated streptavidin (Amersham Life Science, Piscataway, NJ) was added to each well and incubated for 1 hour at room temperature. Finally, p-nitrophenyl phosphate disodium (1 mg/ml in 10 mM diethanolamine [pH 9.5] containing 0.5 mM MgCl2; 50 μl/well) (Sigma) was added as substrate for alkaline phosphatase, and the absorbance of each well was measured at λ405 using a microtiter plate reader (E1×808; BioTek Instruments, Winooski, VT).
Immunization of mice with recombinant human VG1 and AG1.
To examine whether, as in the case of AG1, immunity to VG1 can induce arthritis, age-matched male and female BALB/c mice were immunized with recombinant VG1 or AG1 with complete and incomplete adjuvants or with adjuvants only (control), using a protocol described elsewhere (5, 6) with some modifications. Briefly, on day 0, mice were immunized intraperitoneally with 50 μg AG1 or VG1 (1 mg/ml in PBS) emulsified at a 1:1 (volume/volume) ratio with Freund's complete adjuvant (CFA; Life Technologies) or with PBS and CFA alone. On days 16, 37, 56, and 81, mice were again immunized with 50 μg AG1 or VG1 (1 mg/ml in PBS) emulsified with Freund's incomplete adjuvant (IFA; Life Technologies) or with PBS and IFA alone. In the second experiment, female mice were immunized with recombinant VG1, using the same protocol as in the first experiment. Animals were examined daily for joint redness and swelling (paws and knees) indicative of polyarthritis. The severity of polyarthritis was scored as previously described (7). Animals were killed, according to institutional animal care protocols, at the times indicated in Table 1.
|Antigen, no. of mice/sex||Incidence of polyarthritis, no. (%)||Severity of arthritis, (0–16)||Onset of arthritis, day||Histologic abnormality, no. (%)|
|AG1, 15/F||15 (100)||9.8||80||67–110||8 (53)||5 (33)|
|AG1, 15/M||13 (87)||6.3||100||67–128||3 (20)||4 (27)|
|VG1, 20/F||2 (10)||2||158||149–168||8 (40)||3 (15)|
|VG1, 25/M||1 (4)||1||149||149||8 (32)||2 (8)|
|Adjuvant, 5/F||0 (0)||0||No arthritis||NA||0 (0)||0 (0)|
|VG1, 13/F||0 (0)||0||No arthritis||NA||6 (46)||5 (38)|
|Adjuvant, 13/F||0 (0)||0||No arthritis||NA||0 (0)||0 (0)|
Axial skeletons were fixed in 10% buffered formalin and decalcified in 10% formic acid. Paraffin-embedded sections (6 μm thick) were stained with hematoxylin and eosin, by standard techniques (5).
Immunolocalization of versican.
Versican was localized immunohistochemically, using monoclonal mouse IgG1 antibody (mAb) 12C5 (Developmental Studies Hybridoma Bank, Department of Biological Sciences, University of Iowa, Iowa City), which recognizes an epitope located in the G1 region of versican in humans and rats (19). As a control, IgG1 antibody 2E10H2 (American Type Culture Collection, Rockville, MD), which is directed against canine adenovirus type I, was used. The method of localization involved fixation of intact knees and spines from adult Lewis rats (skin and excess muscle removed) in 4% paraformaldehyde in PBS for 2 hours at room temperature. After rinsing in PBS and in 50 mM Tris HCl (pH 7.2) overnight at 4°C, tissues were decalcified in 0.5M EDTA at 4°C. Frozen sections (6 μm) were prepared and again fixed for 5 minutes in 4% paraformaldehyde in PBS. They were treated with chondroitinase ABC to remove chondroitin sulfate and then stained with diluted mAb for 30 minutes at room temperature. After rinsing, biotinylated rabbit anti-mouse IgG was used, followed by application of streptavidin-peroxidase (Dako, Glostrup, Denmark). Color development and other details are described elsewhere (20).
Isolation of antigen-specific T lymphocyte lines and T cell proliferation assay.
T cell lines were generated from draining popliteal lymph nodes from VG1- or AG1-immunized mice as previously described (15). Briefly, mouse paws were injected with an emulsion of antigen and CFA. These T cell lines were phenotyped using a fluorescein isothiocyanate (FITC)–conjugated anti-mouse CD3 mAb (17A2), and an FITC-conjugated anti-mouse CD4 mAb (L3T4) (PharMingen, Mississauga, Ontario, Canada). Cells were analyzed on a Coulter Epics XL flow cytometer (Beckman Coulter, Montreal, Quebec, Canada). A standard proliferation assay (15) was used to determine responses to synthetic overlapping peptides representing the sequences of human VG1 and AG1 and selected mouse sequences (Figure 1). Results were expressed as a stimulation index (SI) (15). The peptides capable of initiating a significant T cell response (SI ≥5.0) were identified as containing T cell epitope(s).
Enzyme-linked immunosorbent assays (ELISAs) for antibody reactivity and isotyping.
Blood samples were collected by saphenous vein bleeding, 14 days after the fifth injection of recombinant VG1 or AG1. Using recombinant VG1 and AG1, overlapping synthetic peptides of human AG1 and VG1 which represent the sequences of each G1 domain, and selected mouse peptide sequences (Figure 1), we examined immunoreactivity with diluted sera (1/400) by ELISA, as previously described (15, 21). Antibodies were isotyped using a secondary mouse MonoAB ID Kit (Zymed, South San Francisco, CA).
Immunoaffinity purification of antibodies.
VG1- and AG1-specific antibodies were purified from sera collected from VG1- and AG1-hyperimmunized BALB/c mice, using antigen-immobilized agarose gel chromatography. The affinity columns were prepared using recombinant AG1 and VG1 and an AminoLink Plus Immobilization Kit (Pierce, Rockford, IL), and antibodies were isolated according to the manufacturer's instructions.
Western blotting of VG1- and AG1-specific mouse antibodies with mouse versican and aggrecan.
PGs were extracted from mouse spines by treatment overnight at 4°C with 10 volumes of 4M guanidinium chloride and treated with chondroitinase ABC according to the protocol of the manufacturer (ICN Biomedicals, Irvine, CA). Samples were electrophoresed in 8% SDS-PAGE gels under nonreducing conditions. They were examined using Coomassie blue staining and by Western blotting, as previously described (7). Membrane blots were probed using affinity-purified mouse antibodies to VG1 and AG1 and with the versican-specific mouse mAb 12C5. Enhanced chemiluminescence was used to detect antibody binding, with a kit from Amersham Pharmacia Biotech (Baie d'Urfé, Quebec, Canada).
Characterization of recombinant human VG1 and AG1.
Recombinant human VG1 and AG1 were expressed in High Five insect cells with a yield in the range of 1–2 mg/liter. Silver staining after electrophoresis revealed that the recombinant VG1 and AG1 contained 2 or more bands (Figure 2). Both AG1 and VG1 contained a single amino terminus. Each G1 domain band reacted with the carboxyl-terminal His-tag sequence antibody and with the specific antibody to versican (12C5) or aggrecan (IC6) on Western blotting (results not shown). The individual heterogeneity of each protein thus must result from differences in glycosylation. In the case of VG1 there are 2 potential glycosylation sites and in AG1 there are 7 sites that can be glycosylated, in this expression system where variable glycosylation is a known feature (22). This increase in glycosylation in AG1 would account for the increased number of bands seen with this protein. The authenticity of the recombinant human VG1 and AG1 proteins was confirmed by full-length nucleotide sequencing of the constructs (McGill University Core Facility) and by amino acid sequencing of the N-terminal 20 residues of the purified proteins (Shriners Hospital for Children Core Facility) (results not shown). Both had sequences identical to the published sequences for human aggrecan (23) or human versican (24).
The native globular domains of versican and aggrecan are known to bind to HA. Recombinant VG1 and AG1 were examined for binding to HA, which is dependent upon a native conformation. Both VG1 and AG1 bound directly to biotinylated HA and competitively with unlabeled HA (results not shown), indicating that they are folded in a native conformation.
Presence of versican in the intervertebral disc and SI joints.
Immunolocalization revealed that versican was concentrated in distinct sites in the spine and SI joints of the rat. It was found in the annulus, but not in the nucleus or the end-plate (Figures 3A–C). It was present in the ligaments of the spine and their entheseal insertion sites with bone (Figure 3B). No staining was seen in the control (Figure 3D). Versican was also seen in association with the SI joint in entheses (Figures 3E and F) and hyaline cartilage (Figure 3E). It was not detectable in the articular cartilage of the knee (Figure 3G), staining being similar to that in the control (Figure 3H). These results provide an indication of the probable distribution of versican in the mouse.
Unlike recombinant human AG1, recombinant human VG1 did not induce significant peripheral arthritis in BALB/c mice (Table 1). In experiment 1, repeated injection of male and female BALB/c mice with recombinant human AG1 produced a clinically evident peripheral arthritis (with a mean severity score of 9.8 in females and 6.3 in males) and an average time until onset of 80 days in females and 100 days in males. Arthritis was seen in all females and 87% of males (Table 1). Similar sex differences were observed previously in relation to aggrecan (5). In contrast, animals injected with VG1 failed to demonstrate any clear evidence of joint swelling. There was, however, mild redness of the paws in 10% of females and 4% of males, with a severity score between 1 and 2 and an average time to onset of 149 days (Table 1). Arthritis was not seen in any of the control animals injected with adjuvant alone (Table 1). In the second experiment (Table 1), only female mice were immunized with recombinant VG1; animals were killed on day 218 and no peripheral arthritis was observed at that time.
Histologic examination of the lumbar spines revealed that both recombinant VG1 and recombinant AG1 induced spondylitis in BALB/c mice. With recombinant AG1, the findings were similar to those reported for aggrecan (5) and native AG1 purified from bovine cartilage (7). In experiment 1 (Table 1), 53% of the female AG1-injected mice and 20% of the male mice developed spondylitis of the kind described previously (7). Immunization with adjuvant alone failed to induce spondylitis. VG1 injection resulted in the induction of spondylitis in 40% of the female mice and 32% of the males. In experiment 2, spondylitis was observed in 46% of female VG1-injected mice but in none of the control mice. Early mononuclear cell infiltration was seen in the ligamentous tissue and entheses adjacent to the annulus (Figures 4B and C), where versican is concentrated in the rat. This pathology was never observed in control animals injected only with adjuvant (Figure 4A). In more advanced disease, loss of structure of the intervertebral disc and erosion of bone were observed in association with this mononuclear cell infiltration (Figures 4D and E). In these cases, the nucleus was frequently destroyed and the annulus sometimes exhibited chondromatosis, with cartilaginous tissue replacing the once-fibrous annulus (results not shown).
Sacroiliitis and enthesitis.
In addition to spondylitis, we observed sacroiliitis in mice injected with AG1 or VG1. This had not previously been documented for AG1. In experiment 1, this was seen with AG1 treatment in 33% of females and 27% of males. VG1 injection induced sacroiliitis in 15% of females and 8% of males. In experiment 2, the incidence was 38% in females. Thus, there was a greater incidence of both spondylitis and sacroiliitis with VG1 in experiment 2, in which only female mice were examined. Mononuclear cell infiltration was seen in the entheses of the SI joint (Figures 5A–C); this was absent in control mice treated with adjuvant only (Figures 5D–F).
Interrelationships between spondylitis and sacroiliitis.
The onset of spondylitis did not necessarily correlate with SI involvement. For example, of the female AG1-injected mice in experiment 1, only 20% had both spine and SI joint involvement, 33% had only spine involvement and 13% had only SI disease. Of the male mice, only 7% had only spine involvement, 13% had only SI joint involvement, and 13% had both spine and SI joint involvement. Among VG1-injected mice in experiment 1, the SI involvement, when present, was usually seen only in mice with spondylitis. The majority of mice had only spondylitis, regardless of sex. In experiment 2, 15% of the females showed only spine involvement, 8% showed only SI joint involvement, and 31% had both spine and SI joint involvement.
T cell immunity.
We established 3 VG1-specific and 3 AG1-specific T cell lines from draining lymph nodes of immunized animals. All of the T cell lines were CD4+ (>90% CD3+ and CD4+ cells) and exhibited specificity for the immunizing proteins. Examples of 2 representative T cell lines are shown in Figure 6. For VG1 T cell lines, a dominant T cell epitope was identified in peptide V7 (Figure 6A), corresponding to amino acid residues 91–115 located in the amino-terminal loop of VG1. The dominant T cell epitope of human AG1–specific T cell lines was located in peptide A5 (Figure 6C), corresponding to amino acid residues 58–81 (Figure 1), located in the amino-terminal region of AG1. This region includes the dominant T cell epitope of bovine AG1 that we previously identified (21). There was no reactivity to purified protein derivative (Figure 6). The VG1-specific T cell line showed a low but significant response (SI 7) to human AG1, but not to peptides thereof (Figure 6B). The AG1-specific T cell line also revealed a reduced but significant cross-reactivity with human VG1 (SI 6.5), but not to peptides thereof (Figure 6D). These specificities shown in Figure 6 were observed for all VG1 and AG1 T cell lines.
We have previously shown that the dominant T cell epitope of bovine AG1, which contains the sequence shown in human sequence peptide A5, cross-reacts with the homologous sequences in human and mouse AG1 (21). In the present study we confirmed that T cells that recognize the human A5 sequence of AG1 cross-react with the corresponding mouse peptide containing this sequence (data not shown). In the case of VG1-specific T cell lines that recognized peptide V7, there was no evidence of cross-reactivity to the corresponding mouse sequence in peptide MV1 (data not shown).
B cell immunity.
B cell immunity to VG1 and AG1 was examined by analyses of sera by ELISA. Sera from AG1-immunized mice equally recognized both AG1 (Figure 7A) and VG1 (Figure 7C), while sera from VG1-injected mice showed a preferential response to VG1 (Figure 7B) but also reacted with AG1 (Figure 7D). There was limited reactivity to a His-tagged peptide in all mice (Figure 7). Sera from mice treated with adjuvant only did not react with VG1 or AG1, or with peptides thereof (data not shown).
Two major regions in AG1 were recognized by sera from AG1-injected mice, one at residues 148–216 (peptides A11–A14) and the second at residues 268–306 (peptides A19, A20). Both regions are within the HA binding domains, where there is the most interspecies sequence homology. For example, residues 268–306 have only a single mismatch at residue 273 between mouse and human (glycine in mouse instead of alanine in human) (23, 25). In contrast, VG1-injected mice recognized 4 regions in VG1: at residues 1–25 (peptide V1), residues 151–190 (peptides V11, V12), residues 256–295 (peptides V18, V19), and residues 301–343 (peptides V21–V23). The first region is more variable, showing a 7–amino acid mismatch between human and mouse sequences (23, 25). The second and third regions (VG1151–190 and VG1256–295) are the corresponding counterparts of B cell–reactive regions in AG1 (AG1148–216 and AG1268–306, respectively). The fourth B cell region is located at the C-terminal end of VG1 and shows a 6–amino acid mismatch between human and mouse sequences (24, 26).
Sera of both AG1-treated and VG1-treated mice showed significant cross-reactivity, not only at the protein level but also at the peptide level, to domains in the most conserved region of both molecules. Human VG1–injected mice recognized the homologous domain of AG1 represented by peptides A12 and A13, despite a 2–amino acid mismatch between human AG1 and VG1 (Figures 7B and D). Also, human AG1–immunized mice recognized the homologous sequence of AG1 represented by peptides V11 and V12 (Figures 7A and C).
We were interested to determine whether such cross-reactivity existed between the homologous human and mouse domains, because a multiple sequence alignment showed that human AG1163–194 and mouse AG1163–194 are identical, and there is only a 2-residue mismatch between human VG1159–190 and mouse VG1159–190, at residues 161 and 182 (Figure 8A). To address this issue, 2 mouse VG1 overlapping peptides (MV2 and MV3) corresponding to mouse VG1159–190 and 2 mouse AG1 overlapping peptides (MA1 and MA2) corresponding to mouse AG1163–194 were synthesized and examined for a cross-reactivity. The results clearly showed that sera from both human AG1– and human VG1–immunized mice recognized VG1 and AG1 and the homologous mouse AG1 and VG1 sequences (Figures 8A and B). There was no reactivity to unrelated peptides A4 and V4.
Nature of immunoglobulins reactive with VG1 and AG1.
To determine the type of antibodies to VG1, AG1, VG1 peptides (V11 and V12), and AG1 peptides (A12 and A13) in the sera of the mice hyperimmunized with human VG1 and AG1, antibodies were isotyped. Binding to recombinant proteins (VG1 and AG1) revealed that the majority of reactive antibodies belong to IgG1, IgG2a, and IgG2b with a κ light chain, while IgG3, IgA, and IgM contributed less to the immunity (Figure 9A). When the dominant peptide epitopes were used, the immunoglobulin pattern was unchanged, except that IgM reactivity was clearly observed (Figure 9B).
Recognition of mouse versican and aggrecan by affinity-purified antibodies in mice hyperimmunized with recombinant human VG1 and AG1.
PGs were extracted from spines of healthy mice to determine whether antibodies in arthritic mice reacted with these mouse PGs. Figure 10 shows that affinity-purified antibodies to human VG1 and AG1 isolated from the sera of these hyperimmunized mice cross-reacted with these recombinant proteins and with fragments of VG1 and AG1 present in spine extracts, which were distinguished by their different molecular sizes and their reactivity to anti-versican antibody 12C5 and to an anti-aggrecan antibody (results not shown). The 2 main AG1-reactive fragments were similar in size, the larger corresponding to the migration position of AG1. The smaller, faster-moving fragment reacted with the anti-VG1 antibody and with the versican-specific antibody 12C5. In view of the sizes of these fragments, the primary VG1- and AG1-reactive species in extracts are clearly degradation products of versican and aggrecan. Higher molecular weight species seen with all antibodies no doubt reflect larger, less-degraded fragments of these molecules.
Our earlier studies have revealed that the G1 globular domain of aggrecan, AG1, appears to be the focus of immunity to this molecule that results in the induction of polyarthritis and spondylitis in BALB/c mice (7). Since the G1 domains of aggrecan and versican are of similar structure, we anticipated that the G1 globular domain of versican, VG1, might also induce this type of arthritis. The results of this study show that immunity to the G1 domain of versican can indeed induce the type of spinal and SI pathology seen in spondylarthropathies such as AS, but without the polyarthritis seen with AG1 or aggrecan.
Versican is a member of the family of proteoglycans, which includes aggrecan. They bind to HA through the G1 globular domain. Versican has 4 isoforms (V0, V1, V2, and V3) that result from differential splicing (27). All isoforms share a common HA binding G1 domain which contains 2 HA binding molecules. The sequence and the structure of the first 350 residues from the N-terminus of human VG1 exhibit considerable homology with human AG1 and link protein.
We show here that versican is enriched in the annulus of the rat intervertebral disc, in contrast to its absence from the nucleus and end-plate of the disc and from articular cartilage, where aggrecan is concentrated. Others have reported the presence of versican in the nucleus, although it is mainly concentrated in the annulus (28). Its relative absence from hyaline articular cartilage confirms the findings of an earlier study, in which it was seen only at pericellular sites in adult human cartilage (29). We did, however, detect versican in the hyaline cartilage of the rodent SI joint. Its presence in ligaments and entheses is also consistent with an earlier report (30). The distribution of versican as observed in the rat therefore corresponds to sites where histopathologic abnormalities were observed in mice immunized with VG1. This distribution of versican may therefore explain, in part, the general lack of development of a recognizable polyarthritis in contrast to observations with aggrecan, which, unlike versican, is a principal component of hyaline cartilage. As in the case of VG1, immunity to AG1 was observed at sites where aggrecan is detected, such as the articular cartilage, nucleus of the intervertebral disc, and the SI joint.
The pathology we observed in BALB/c mice closely resembles that seen in AS in humans (30). Thus, the SI joint and spine exhibited clear features of an early enthesitis in the same sites where versican was localized. Importantly, the enthesitis was also of the kind observed in patients with AS (31). Both with VG1 and with AG1 treatment, the incidence of spondylitis and sacroiliitis was higher in females, as previously observed with aggrecan (5). In future studies we plan to investigate whether, in more advanced disease, VG1 induces the formation and ossification of spinal syndesmophytes (which are seen in AS patients) and to seek more evidence of enthesitis at sites of insertion of ligaments and tendons into bone.
Analysis of the immune responses to AG1 and VG1 revealed evidence of limited T cell cross-reactivity between recombinant human VG1 and recombinant human AG1, and vice versa. There was no evidence of cross-reactive T cell immunity between human VG1 and mouse VG1 at the peptide level. Previously we showed that CD4+ T lymphocytes reactive with bovine AG1 can cross-react with the corresponding sequences in human and mouse AG1 (21). In the present study we confirmed that T lymphocytes that recognize the same human sequence (contained in peptide A5) cross-react with the corresponding mouse sequence. In contrast to the lack of T cell cross-reactivity with mouse VG1, there was clear evidence of cross-reactive B cell immunity between human VG1 and homologous sequences in human AG1 and in mouse VG1 and AG1, and with degradation products of mouse versican and aggrecan containing the G1 domains. Antibodies to human AG1 also cross-reacted with fragments of mouse versican and aggrecan.
In AG1, there are 2 B cell–dominant regions within the most conserved areas. In VG1, there are 4 B cell–dominant regions. The first and fourth regions are located in more variable parts of VG1. The second and third regions are homologous counterparts of AG1 and are probably the cross-reactive and arthritogenic epitopes. Interestingly, these regions are involved in HA binding. In aggregates, the HA binding regions are normally not recognized by the immune system, presumably because of the presence of HA. But when BALB/c mice were immunized with AG1 or VG1 alone, the absence of HA no doubt favored exposure of the HA binding regions, allowing the immune system to recognize them as nonself. Such a situation may be similar to that in patients with arthritis. During arthritis development, the G1 domain is released into the synovial fluid and is often denatured, being unable to bind HA (32). This would favor immunity to the G1 domain. Similarly, the antibodies to VG1 and AG1 reacted with degradation products in mouse spine extracts of similar sizes to the G1 domains of these molecules.
Both recombinant human VG1 and recombinant human AG1 bind to HA. Differences in binding could conceivably influence T cell immunity to these molecules and may in part explain the histopathologic differences we observed. In separate studies, we have performed binding studies using the BIACORE system to show that VG1 binds more strongly than AG1 to HA (Shui S, Zhang Y, Poole AR, Mort JS: unpublished observations). These binding differences may influence the development of histopathologic abnormalities by affecting neoepitope exposure during the immune response. This remains to be investigated in future studies.
In view of the lack of T cell cross-reactivity to mouse versican, it would appear likely that cross-reactive B cell immunity between human VG1 and AG1 and between mouse VG1 and AG1 plays an important role in the development of the pathology induced by VG1 and AG1. Such autoimmunity would specifically favor the development of disease. The lack of cross-reactive T cell immunity between human VG1 and mouse VG1 and the absence of versican in hyaline cartilage of knee joints may help explain the general lack of a well-defined polyarthritis in VG1-immunized mice, since in the case of immunity to aggrecan, the development of this polyarthritis is dependent upon the involvement of CD4+ T cells that are reactive to injected bovine or human AG1, which cross-react with mouse AG1 (6, 33).
Taken together, these observations provide important new evidence suggesting that humoral immunity to these G1 domains of versican and aggrecan and the observed cross-reactivities may play important roles in the development of the abnormalities observed in experimental animal models of spondylarthropathies. It remains to be established whether humoral immunity of this kind can alone induce such abnormalities in human patients with these arthropathies.
We thank Virabhadrachari Vipparti, MSc, and Zhi-Hua Shi, MSc, for assistance in the preparation of recombinant proteins, Elisa de Miguel for peptide synthesis, HPLC purification, and protein sequencing, and Mark Lepik and Guylaine Bedard for art work. We are grateful to John Mort, PhD, and Peter Roughley, PhD, for their constructive comments.
- 1Ankylosing spondylitis: clinical aspects. In: TaurogJD, CalinA, editors. The spondyloarthritides. Oxford (UK): Oxford University Press; 1998. p. 27–40..
- 7Immunity to the G1 globular domain of the cartilage proteoglycan aggrecan can induce inflammatory erosive polyarthritis and spondylitis in BALB/c mice but immunity to G1 is inhibited by covalently bound keratan sulfate in vitro and in vivo. J Clin Invest 1996; 97: 621–32., , , , , , et al.
- 31Pathology of the spondylarthropathies. In: CalinA, editor. Spondyloarthropathies. Orlando: Grune & Stratton; 1984. p. 43–68..