To characterize experimentally induced spondylarthropathy (SpA) in arthritis-susceptible inbred mice and in their F1 and F2 hybrid generations of susceptible and resistant mouse strains.
To characterize experimentally induced spondylarthropathy (SpA) in arthritis-susceptible inbred mice and in their F1 and F2 hybrid generations of susceptible and resistant mouse strains.
SpA was induced in susceptible BALB/c and C3H/HeJCr (C3H) strains of mice, and in their F1 and F2 generations derived from intercrosses with arthritis- and/or spondylitis-resistant DBA/2 and DBA/1 parent strains, by systemic immunization with cartilage proteoglycan (PG) aggrecan. The incidence and severity of PG-induced spondylitis (PGIS) were scored histologically, and these scores for spine involvement were correlated with serum antibody and cytokine levels and with in vitro T cell responses to cartilage PG.
PGIS was induced by systemic immunization with cartilage PG in adjuvant, and ∼60–70% of susceptible mouse strains and their F2 hybrids developed spondylitis either with or without arthritis. Adjuvants, particularly those activating the innate immune system and enforcing the Th1 dominance, had significant effects on the outcome and progression of SpA. The DBA/1 strain appeared to carry genes protecting this strain and its F1 and F2 hybrids from spondylitis, whereas the DBA/2 strain, although resistant to PGIS, harbored genes permitting PGIS in its hybrid generations. Arthritis- and/or spondylitis-susceptible BALB/c and C3H parent strains and their F2 hybrids exhibited the highest incidence and severity of spondylitis.
PGIS, a murine model of autoimmune spondylitis, shows similarities to ankylosing spondylitis. Segregation of susceptibility to PG-induced arthritis (PGIA) from that to PGIS in different genetic crosses suggests that PGIA and PGIS are separate diseases. Therefore, this model allows for the elucidation of genetic components involved in the etiology of SpA, independent of those controlling the susceptibility to PGIA.
Spontaneous or experimentally induced spondylarthropathy (SpA) has been reported, to date, in only a few animal models (1–4), and autoimmune mechanisms of SpA are thought to be involved in only HLA–B27–transgenic rodents (5–7) and mice with proteoglycan (PG)–induced arthritis (PGIA) (8–10). Because HLA–B27–transgenic animals, when maintained in germ-free conditions, do not develop spondylitis (7, 11), the molecular mimicry of bacterial antigens is believed to be a contributing factor to the pathomechanisms of SpA in transgenic rodents and, perhaps, in humans as well. Although neither the HLA–B27–transgenic model nor the PGIA model is identical to human ankylosing spondylitis (AS), both models mimic genetic and/or pathologic abnormalities present in the human disease.
Immunization of BALB/c or C3H/HeJCr (C3H) mice with cartilage PG induces progressive polyarthritis (PGIA), which is frequently accompanied by spondylitis (8, 10). Since PGIA reaches 100% incidence in these 2 susceptible murine strains, and because the onset of spondylitis cannot be precisely determined in vivo in mice, involvement of the spine has been considered a concomitant symptom of PGIA. However, our recent study suggested that PGIA and PG-induced spondylitis (PGIS) could be 2 independent diseases (12), even though both are induced by immunization with cartilage PG. The major supporting argument for this notion is that, in a mixed genetic background, PGIA occurs without spondylitis, and a number of intercrossed mice with PGIS do not have peripheral arthritis.
The goal of the present study was 2-fold. First, to investigate the relationship between spine involvement and peripheral joint involvement, we analyzed the occurrence and incidence of PGIS in arthritis-susceptible and arthritis-resistant murine strains. Second, to gain insight into the genetic control of spine involvement, we used F2 hybrids of PGIS-susceptible and PGIS-resistant strains to determine the incidence and severity of SpA in the context of disease-related parameters of inflammation and the immune response.
Human cartilage was obtained from patients undergoing joint replacement surgery. The use of this cartilage for antigen (PG) isolation was approved by the Institutional Review Board, and all animal experiments were approved by the Institutional Animal Care and Use Committee. Human cartilage PG was used for immunization of 16–26-week-old female and male BALB/c, DBA/2, and C3H mice, which were purchased from the National Cancer Institute (Kingston Colony, NY). Female and male DBA/1J mice were purchased from Jackson Laboratory (Bar Harbor, ME) and immunized with human type II collagen (CII) as previously described (13, 14) at 6–8 weeks of age. F1 and F2 hybrids of these inbred mouse strains were generated in our animal facility (Comparative Research Center), maintained in a pathogen-free, but not germ-free, environment, and immunized exactly the same way as their parent strains. As shown in Table 1, some inbred mice and their F1 and F2 offspring were also studied retrospectively, by using available laboratory data and by examining new histologic sections of formalin-fixed archived spine tissues.
|Strain, adjuvant||MHC allele||Immunization protocol||Incidence of spondylitis||Incidence of arthritis|
|CFA†||H-2d||PGIA||86/130 (66)||127/130 (98)|
|DDA||H-2d||PGIA||38/61 (62)||60/61 (98)|
|CFA||H-2d||PGIA||0/32 (0)||0/32 (0)|
|DDA||H-2d||PGIA||2/50 (4)||0/50 (0)|
|BALB/c × DBA/2|
|DDA||H-2d × H-2d||PGIA||10/31 (32)||0/31 (0)|
|CFA‡||H-2d × H-2d||PGIA||48/160 (30)||38/160 (24)|
|DDA||H-2d × H-2d||PGIA||137/223 (61)||97/223 (43)|
|CFA||H-2q||CIA||0/42 (0)||41/42 (98)|
|DDA||H-2q||CIA||0/40 (0)||40/40 (100)|
|DDA||H-2q||PGIA||0/24 (0)||0/24 (0)|
|BALB/c × DBA/1|
|CFA||H-2d × H-2q||PGIA||0/19 (0)||8/19 (42)|
|CFA‡||H-2d × H-2q||PGIA||0/102 (0)||32/102 (31)|
|DDA||H-2d × H-2q||PGIA||0/61 (0)||23/61 (38)|
|CFA‡||H-2d × H-2q||CIA||0/115 (0)||45/115 (39)|
|CFA‡||H-2k||PGIA||26/38 (68)||36/38 (95)|
|DDA||H-2d × H-2k||PGIA||148/212 (70)||185/212 (87)|
|BALB/c × C3H|
|DDA||H-2k||PGIA||24/36 (67)||36/36 (100)|
Purification of cartilage PG aggrecan (henceforth referred to as PG), isolation of CII, and immunization protocols were the same as described previously (8, 13, 15, 16). As a standard method for immunization with PG, the first antigen injection (100 μg PG protein) was administered intraperitoneally (IP) in Freund's complete adjuvant (CFA) (Difco, Detroit, MI), and the same doses of antigen were injected IP as second and third booster injections in Freund's incomplete adjuvant on weeks 3 and 6. Alternatively, 100 μg PG protein was injected IP with a synthetic adjuvant, dimethyldioctadecyl–ammonium bromide (DDA) (2 mg/100 μl), to avoid the harmful side effects of CFA (15–17). For immunization with CII, the CII was dissolved in 0.1M acetic acid, diluted in phosphate buffered saline. One hundred micrograms of CII in 100 μl was then emulsified with an equal volume of CFA or DDA and injected intradermally into the base of the tail on days 0 and 21. Animals immunized with cartilage PG received a total of 4 injections (except the C3H mice), whereas those immunized with human CII received a third antigen injection with the addition of either CFA or DDA as adjuvant. All animals, except for those in the first group listed in Table 1 (12), were killed at 98–126 days after the first injection.
The onset and severity of arthritis were determined using a visual scoring system that was based on the extent of swelling and redness of the paws (8, 13, 15, 16). Animals were inspected weekly during the first 3 weeks and then (after the second injection) 3 times per week. The degree of joint swelling for each paw (scores ranging from 0 to 4) was expressed as a cumulative arthritis score, with a possible maximum severity index of 16 per animal. For scoring of arthritis onset, the first clinical appearance of swelling was defined as the onset, and an empirical onset score ranging from 6 (the earliest date) to 0 (end of experiment) was established over a 60-day period from day 10 after the second PG injection (i.e., from day 31 of the experiment) (14, 17, 18).
SpA was confirmed by radiography of the spines and sacroiliac joints of PG-immunized mice using a Hewlett Packard Faxitron instrument (18 seconds, 65 kV; Buffalo Grove, IL) and high-resolution film (Kodak X-omat TL, Rochester, NY) (Figures 1a–d). After acquisition of radiographic images, spines were decalcified and embedded in paraffin, and light microscopy scoring of hematoxylin and eosin–stained sections was performed as described previously (12). In brief, spondyldiscitis, i.e., inflammatory cell infiltration around the intervertebral disk (IVD), was recorded as a severity score of 1, while <50% resorption of the IVD received a score of 2, advanced 50–100% resorption of the IVD was recorded as a score of 3 (Figure 1e), and cartilaginous/bony ankylosis of neighboring vertebral bodies was given a score of 4. An average of 18.7 IVDs per mouse was scored histologically (Table 1), which included the distal cervical (from C3/C4) segment and all thoracic and proximal lumbar (to L3/L4) IVDs. Finally, a spondylitis index was calculated for each animal, by dividing the cumulative score by the number of IVDs examined.
Serum amyloid A (SAA) (an acute-phase protein in mice) and serum cytokine levels (tumor necrosis factor α [TNFα], interleukin-1β [IL-1β], IL-4, and IL-6) were measured using enzyme-linked immunosorbent assay (ELISA), with paired antibodies purchased from BD Biosciences (San Diego, CA) or R&D Systems (Minneapolis, MN). Serum levels of PG-specific antibodies were measured by ELISA as described previously (13, 15, 16, 18).
Statistical analysis was performed using SPSS software, version 10.0 (SPSS, Chicago, IL). Student's t-test was used to compare the results of 2 groups. Spearman's rho test was used to determine correlation coefficients. P values of less than 0.05 were considered significant.
Incidence and severity of PGIA and PGIS in inbred mice and their F1and F2hybrids. As has been reported previously, 97–100% of PG-immunized BALB/c and C3H mice developed peripheral arthritis at 2 weeks after the third antigen injection (8, 10, 17). Massive inflammatory cell infiltration, pannus formation, and cartilage and bone erosion characterized the histopathologic picture of the affected joints. None of the DBA/1 or DBA/2 parents nor the (BALB/c × DBA/2)F1 hybrids developed arthritis until the end of the 14–18-week experimental period (Table 1).
Whole-body radiographs revealed axial involvement (initial narrowing of the sacroiliac joint or intervertebral space) at ∼8 weeks after the third PG + CFA injection (Figures 1A–D) and as early as 3–4 weeks after the third PG + DDA injection. Thus, DDA accelerated the onset of PGIS, but the spondylitis still developed several weeks later than the peripheral arthritis. Importantly, SpA was detected in only PG-immunized, but not in CII-immunized, mice, and abnormalities could not be detected in any of the control (arthritis-resistant) animals/strains (Table 1) (12, 13, 15). As described previously (12), cartilage surface erosions were first observed in the sacroiliac joints. PGIS was progressive and very similar in all affected animals/strains (Table 1), and radiographic abnormalities of the spine and sacroiliac joints were supported by the histologic observations. During the course of the spine disease, the lumbar and then the proximal thoracic and distal cervical segments became involved (Figure 1E), but not all IVDs were equally affected at any given time point. Some IVDs seemed to be intact or mildly damaged even when the vast majority of the disks were resorbed and the vertebral bodies underwent ossification or fusion (12).
The incidence of PGIS was 62–70% in inbred BALB/c and C3H mice, whereas only weak and sporadic discitis (involvement of only 1–2 IVDs) occurred in 2 (4%) of 50 PG + DDA–immunized DBA/2 mice. Moreover, no spondylitis was found in PG- or human CII–immunized animals of the DBA/1 strain (Table 1). The incidence and severity of spondylitis were highly comparable in both PGIS-susceptible inbred strains (BALB/c and C3H), regardless of the adjuvant used (CFA or DDA); nevertheless, as described above, PG immunization with DDA as adjuvant resulted in an earlier onset of PGIS.
Although F1 hybrids of the BALB/c × DBA/2 intercross were fully resistant to PGIA, unexpectedly, more than 30% of them developed PGIS, whereas none of the F1 hybrids of BALB/c × DBA/1 developed PGIS. Even more surprising was that F2 hybrids of BALB/c × DBA/2 (PGIS-susceptible versus PGIS-resistant parent strains, respectively) and (BALB/c × C3H)F2 mice (both of which have PGIS-susceptible parent strains) showed a similarly high incidence (63–70%) and severity of spondylitis (Table 1 and Figure 2), but none of the F2 hybrids of the BALB/c × DBA/1 intercross developed spine involvement. This finding was unexpected, because the F2 hybrids of both genetic intercrosses (BALB/c × DBA/2 and BALB/c × DBA/1) exhibited a similar incidence (Table 1) and severity of PGIA when immunized using the same protocol. Since the F2 hybrids of major histocompatibility complex (MHC)–matched mice (BALB/c × DBA/2; both having the H-2d haplotype) and MHC-unmatched mice (BALB/c H-2d versus C3H H-2k) exhibited a similarly high incidence and severity of PGIS (Table 1 and Figure 2), we focused, in subsequent studies, on the PGIS in these 2 intercrosses.
Characterization of PGIS in F2hybrids of BALB/c × DBA/2 and BALB/c × C3H mice. Although the onset of arthritis (PGIA) was earlier and the severity was higher in (BALB/c × C3H)F2 hybrid mice than in (BALB/c × DBA/2)F2 mice (Figure 2), the incidence of PGIA (Table 1) and severity of spondylitis (Figure 2) were comparable in the 2 hybrid groups. We found that 137 of 223 (BALB/c × DBA/2)F2 mice (61.4%) and 148 of 212 (BALB/c × C3H)F2 mice (69.8%) developed PGIS (Figure 3A), with a mean ± SD spondylitis (histologic) severity score of 0.69 ± 0.63 and 0.66 ± 0.57, respectively (Figure 2). However, the number of mice considered negative (33.2%) (developing neither spine nor joint involvement) (Figure 3A) and the number that developed only PGIS without arthritis (23.3%) was significantly higher in the (BALB/c × DBA/2)F2 group than in the (BALB/c × C3H)F2 hybrids (9.9% negative and 2.8% with PGIS only) (Figure 3A). Nevertheless, when PGIA and PGIS occurred together (Figure 3A), more progressive arthritis was usually associated with more severe spondylitis (Figures 4A–D).
The next evident question was whether any of the 4 clinical groups (i.e., negative, affected with arthritis only, affected with spondylitis only, or affected by both diseases) could be distinguished based on differences in T or B cell responses and/or disease-related markers measured in the serum. Although the power of statistical significance was reduced because of the high individual differences in F2 hybrid mice (Figure 2) and because of the relatively low number of arthritic (BALB/c × DBA/2)F2 and negative or spondylitis only–affected (BALB/c × C3H)F2 hybrids (Figure 3A), we could still observe significant differences in the serum IL-6 levels when the 4 groups in each of the 2 genetic combinations were compared (Figure 3B). IL-6 was very high in the sera of animals having both arthritis and spondylitis (Figure 3B), and the levels of both serum IL-6 and SAA positively correlated with the severity of PGIS (Figures 4E–H).
We found very strong T and B cell responses to both human (immunizing) and mouse (self) PGs and high cytokine (TNFα, IL-1β, IL-4, and IL-6) concentrations in all PG-immunized mice (results not shown for all cytokines), and compared all possible combinations. Although the anti-PG antibody (both auto- and heteroantibodies), IL-6, and serum TNFα levels showed a significant correlation with the presence of PGIS in (BALB/c × C3H)F2 mice, only the serum IL-6 and SAA levels were significantly higher in (BALB/c × DBA/2)F2 mice with PGIS (Figures 4E and G). However, this difference could be due to the earlier onset and increased severity of arthritis, but was not necessarily related to the spondylitis, in (BALB/c × C3H)F2 mice (Figure 2).
In the present study, we analyzed the relationship between arthritis and spondylitis, both of which can be induced in PGIA-susceptible inbred mouse strains and their F1 and F2 hybrids. Although SpA, as a concomitant phenomenon of PGIA, has been known since the first description of the model (8, 17), this is the first study in which the effects of antigens, adjuvants, and genetic backgrounds have been systematically investigated. We compared not only known SpA-susceptible BALB/c and C3H strains, but also their MHC-matched and MHC-unmatched F1 and F2 generations, which were derived from intercrosses with arthritis (collagen-induced arthritis [CIA])–susceptible (DBA/1) and arthritis-resistant (DBA/2) strains.
The development of PGIS, like PGIA, requires cross-reactive immune (both T and B cell) responses between human cartilage PG used for immunization and mouse (self) PG. IVDs, both nucleus pulposus and annulus fibrosus, contain large amounts of the PG aggrecan (19), which is similar to that in articular cartilage. Since human PG shows very high homology with mouse PG (20, 21), it is not surprising that in mice immunized with human cartilage PG, an immune attack is mounted against cartilaginous tissues in the mouse joints and IVDs. However, this occurs only in special genetic backgrounds, indicating that arthritis susceptibility per se is insufficient for the development of spondylitis. As briefly described, DBA/1 mice develop arthritis (CIA), but not spondylitis, in any experimental condition, i.e., using either males or females, immunization with either cartilage PG or human CII, in either CFA or DDA. Although 30–40% of the F1 and F2 hybrids of the BALB/c × DBA/1 intercross, immunized with either PG or human CII, developed arthritis, we could not detect inflammatory cells around or adjacent to the IVDs in these arthritic or nonarthritic animals when more than 3,500 IVDs of more than 230 spine sections were examined. This observation suggests that the DBA/1 strain carries very strong protective genes against SpA.
Within the DBA/2 strain, which was completely resistant to PGIA and mostly to PGIS as well (i.e., only a small percentage of animals had very mild inflammation adjacent to 1 or 2 IVDs), ∼30% of their F1 hybrids with the PGIS-susceptible BALB/c strain developed spondylitis, with a relatively high severity score (mean ± SD 0.13 ± 0.06). In their F2 hybrids, the incidence (61%) and severity (0.69 ± 0.63) of spondylitis increased further, eventually reaching the values found in the parent BALB/c strain, when both the parental and hybrid mice were immunized with PG in DDA adjuvant (Table 1). Thus, the DBA/2 genome should contain spondylitis susceptibility and protective genes that might be silent in the original background. These observations in the genetic combination of BALB/c and DBA/2, however, are the first proof that PGIA and PGIS possibly represent 2 different diseases (12). BALB/c and DBA/2 strains carry the same H-2d allele. This indicates that the MHC alone (e.g., in DBA/2 mice) is insufficient to control PGIS susceptibility. This notion was also supported by the results in the (BALB/c × DBA/1)F2 generation, in which ∼25% of the immunized mice were homozygous for the H-2d allele (14) but none of the F2 hybrids developed spondylitis (Table 1).
In contrast, when 2 spondylitis-susceptible strains (BALB/c and C3H) were intercrossed, the incidence of PGIS in the F2 hybrids was essentially the same as in any of the 2 parent strains. Thus, it is likely that neither the BALB/c nor the C3H strain has additional genetic loci that could either reduce, as in (BALB/c × DBA/1)F2 hybrids, or increase, as in F2 hybrids of BALB/c × DBA/2 strains, the susceptibility to PGIS. This is not surprising, since the common ancestor of the C3H and BALB/c strains is a female Bagg albino (22). Thus, spondylitis-susceptibility genes could have been present in the Bagg albino before the BALB/c and C3H lines were separated as inbred strains. Because no other inbred strains have, as yet, been found to be susceptible to SpA (9, 13, 17), we believe that only a very few genes control SpA susceptibility, and that these genes are most likely the same in the BALB/c and C3H strains. The first genome-wide screening studies of 223 (BALB/c × DBA/2)F2 hybrid mice have been completed recently, and indeed, only 2 definitive loci (on chromosomes 2 and 18) and 4 suggestive loci (on chromosomes 11, 12, 15, and 19) seem to be linked to PGIS in this MHC-matched generation (23).
Induction of an autoimmune arthritis in genetically susceptible rodents using CFA–based protocols requires at least 1 injection of antigen (either PG or CII) in CFA. This indicates that the activation of the innate immune system with mycobacterial components in oil is as important as the antigen-induced activation of the adaptive immune system. Mycobacterial compounds, such as heat-shock proteins, are known as potent nonspecific cellular stimulators (24), and muramyl-dipeptide (a peptidoglycan) and trehalosedimycolate (a glycolipid equivalent to lipopolysaccharide of Escherichia coli) (15, 25) could play a role in the enhancement of immune reactions to self antigens in autoimmune models. Recently, we found that a hydrophilic/lipophilic quaterner ammonium base, DDA, which incorporates antigens into a liposomal micelle (26), is as effective an adjuvant as CFA (without the side effects of CFA) in the induction of either PGIA or CIA (15). Both CFA and DDA stimulate the innate immune system equally well and also activate antigen-specific effector and regulatory T cells of the Th1 arm of adaptive immunity (15).
We have thus compared the adjuvant effects of CFA and DDA on PGIS. As described earlier (15), DDA together with PG could induce arthritis and spondylitis in BALB/c and C3H mice. In these 2 parent (inbred) strains, we found no difference using either CFA or DDA with PG (Table 1), but DDA appeared to be a more potent adjuvant than CFA in (BALB/c × DBA/2)F2 hybrid mice. We found an ∼2-fold increase in the incidence of both PGIS and PGIA in F2 hybrids of the BALB/c × DBA/2 intercross, when these mice were immunized with PG in DDA instead of PG in CFA (Table 1). However, if a strain was resistant to either PGIA, CIA, or PGIS, the antigen in DDA adjuvant was not sufficient to induce tissue-specific inflammation. In conclusion, although DDA, as a more potent adjuvant than CFA, may influence the incidence of SpA in susceptible mice, the (auto)immune responses to tissue/organ-specific antigen (in this case, PG) and the genetic background (including the appropriate MHC) are the most critical factors in the development of SpA.
The MHC, in general, is the strongest genetic component in autoimmune disorders. The association of HLA–B27 with AS, as evidence of the autoimmune etiology of AS, was first described more than 30 years ago (27). The combination of HLA–B27 with other HLA alleles (HLA–B60 and HLA–B35) was found to increase the genetic predisposition to AS (28, 29), and genome-wide screening studies suggested a polygenic character of the disease (30, 31). Despite intensive research in this area, the pathologic mechanism of AS is unknown. Studies on putative autoantigens implicated the role of molecular mimicry, represented by Klebsiella antigens (32, 33), Yersinia antigens (34, 35), self-recognized HLA–B27 (35–37), or epitopes in cartilage PG (38–41) or versican (42). The present study could not reveal which subdomains within the murine H-2 locus are responsible and/or involved in the genetic predisposition to PGIS in either of the F2 hybrids of BALB/c × C3H or the BALB/c × DBA/2 intercross.
PGIS has been detected in the same inbred strains that are susceptible to PGIA. Nevertheless, the genetic, clinical, and laboratory findings are different in PG-induced arthritis and spondylitis (12). These differences may reflect only quantitative differences in animals affected with both diseases, However, the complete resistance to either arthritis or spondylitis, or both, in different genetic combinations suggests that PGIA and PGIS are 2 distinct diseases. Although many immune system traits and laboratory parameters (including antigen-specific T cell responses, PG-specific IgG isotype ratios, and levels of interferon-γ, IL-1β, or TNFα) could distinguish between arthritic and nonarthritic mice (results not shown), only the high serum IL-6 level seemed to be consistently associated with the presence of SpA in PG-immunized mice. This observation may indicate that PGIS is a more uniform disease than PGIA. Indeed, although several genetic loci have been implicated in different forms of autoimmune/experimental arthritis (17, 43) and in rheumatoid arthritis (43–45), only a very few, probably only 2 dominant, non-MHC loci (on chromosomes 2 and 18) have been identified that could control susceptibility to spondylitis in (BALB/c × DBA/2)F2 hybrid mice (23). Importantly, both of these 2 genetic loci correspond to those identified in human patients with AS (30, 31, 46, 47). Therefore, experimentally induced SpA in mice can serve as a tool for elucidating the genetic, immunomodulatory, and other components that control the development of SpA.
The authors thank Drs. Vyacheslav A. Adarichev, Tamás Bárdos, Andrey B. Nesterovich, and Miklós Tunyogi-Csapó, as well as Deborah J. Hall and Sonja Velins, for providing expert technical assistance.