Cytotoxic T lymphocytes recognize and lyse chondrocytes under inflammatory, but not non-inflammatory conditions

Abstract

The human major histocompatibility complex (MHC) class I allele HLA-B27 is strongly associated with seronegative spondyloarthropathies including ankylosing spondylitis and reactive arthritis. Although of unknown aetiology, one hypothesis suggests that a cytotoxic T cell (CTL) response against a self-antigen at sites of inflammation, such as entheses or joints may be involved. The chondrocyte is one of the major specialized cell types found both in articular cartilage and cartilaginous entheses and therefore is a possible source of such an antigen. CTL recognition of these cells is a potential mechanism for inflammation and cartilage damage, both through direct lysis of chondrocytes and the secretion of pro-inflammatory cytokines such as tumour necrosis factor and interferon-γ (IFN-γ). We test the feasibility of this hypothesis by examining the ability of chondrocytes to present antigen to CTL in vitro. Chondrocytes isolated from the ribcages of mice did not constitutively express detectable levels of MHC class I by fluorescence-activated cell sorting analysis. In addition, they were resistant to lysis by alloreactive and influenza A virus nucleoprotein (NP)-specific CTL. However, treatment of chondrocytes with IFN-γ up-regulated MHC class I expression and rendered the cells susceptible to lysis by CTL. Similarly, IFN-γ-treated chondrocytes infected with influenza A virus were recognized by NP-specific CTL, though with variable efficiency. Thus, we suggest that under certain circumstances CTL-mediated lysis of chondrocytes is potentially a potent mechanism for cartilage damage in vivo, but that low levels of MHC class I on healthy chondrocytes protects from immune recognition in health.

Introduction

The seronegative spondyloarthropathies, including ankylosing spondylitis and reactive arthritis are strongly associated with the human leucocyte antigen (HLA) class I allele, HLA B27. Although the aetiology of these diseases remains obscure, the possibility of autoimmune responses to antigens found in sites of inflammation that are recognized by HLA B27 restricted T cells cannot be ignored. Characteristic features of the spondyloarthropathies include enthesitis (inflammation at a ligament/tendon–bone junction) as well as some cartilage destruction. There are a limited number of cells types in these areas, but amongst the most specialized are chondrocytes which produce tissue specific antigens such as the collagens, cartilage proteoglycans and human cartilage glycoprotein 39 (HCgp39).¹ Cytotoxic T lymphocyte (CTL)-mediated recognition of chondrocytes has the potential to provide a powerful means of cartilage damage and inflammation. Indeed, chondrocyte-specific CTL could contribute to joint damage during arthritis by several pathogenic mechanisms. First, they could directly lyse the chondrocytes, destroying the cells responsible for maintaining cartilage integrity² and causing the release of further autoantigens³ to perpetuate the immune response. Such a mechanism for tissue injury has been postulated in multiple sclerosis, where CTL-mediated lysis of oligodendrocytes by CTL specific for myelin basic protein has been demonstrated in vitro.⁴ Second, autoreactive CTL may be an important source of pro-inflammatory mediators. CTL not only produce chemokines and destructive enzymes,⁵ but after target cell recognition also secrete tumour necrosis factor (TNF), which has long been known to play a central role in the pathogenesis of arthritis through the induction of a pro-inflammatory cytokine cascade.⁶ Nevertheless, with the exception of the seronegative spondyloarthropathies, the potential role of CTL in arthritis has received relatively little attention.

Although a number of studies have demonstrated major histocompatibility complex (MHC) class II restricted T-cell responses to chondrocytes^7,8 MHC class I restricted responses are poorly documented. In this study we have looked at the ability of chondrocytes derived from neonatal mice to be recognized by antigen specific and alloreactive CTL. We show that chondrocytes are relatively resistant to lysis unless they have been induced to up-regulate expression of MHC class I on their surface. They are also able to process and present viral antigen to CTL, but the efficiency of this is less than standard cell lines used for in vitro assays and varies considerably depending on the epitope or presenting haplotype.

Materials and methods

Mice

BALB/c and C57Bl6 mice and F1 crosses, as well as HLA B27 hβ2m⁹, DES T-cell receptor (TCR)¹⁰ and 2C-TCR transgenic mice¹¹ were bred under specific pathogen-free (SPF) conditions at the Institute for Animal Health, Compton, UK.

All animal experiments were performed under a Home Office project License, in compliance with relevant laws and local guidelines, and approved by the Institute for Animal Health Ethical Committee.

Preparation of chondrocytes

Chondrocytes were prepared from the ventral parts of neonatal ribcages based on a method by Lefebvre et al.¹² Ribcages were digested with pronase (Boehringer Mannheim, Mannheim, Germany), 2 mg/ml in phosphate-buffered saline (PBS) for 30 min at 37°, followed by collagenase D (Boehringer Mannheim), 1·5 mg/ml in RPMI-1640 with 10% fetal bovine serum (FBS; R/10), for 60 min at 37°, 5% CO₂. Gentle agitation was applied until soft tissues detached. Cartilages were then washed five times in PBS, covered with fresh collagenase (1·5 mg/ml in R/10) and digested for 5 hr to a single cell suspension. Cells were passed through a strainer to remove remaining bone. Some chondrocyte preparations were incubated with interferon-γ (IFN-γ; Sigma, Poole, UK), 100 µg/ml in R/10 for 48 hr. Cultured cells were re-digested in 0·5 mg/ml collagenase to obtain a cell suspension before use.

Fluorescence-activated cell sorting (FACS) analysis

Cells were blocked in 5% host serum in PBS/1% bovine serum albumin (BSA), stained according to standard protocols and analysed on a FACSCalibur (Becton Dickinson, San Jose, CA). For staining of intracellular antigens 0·3% saponin was added during blocking and staining steps 0·1% saponin during washes.

In vitro restimulation and ⁵¹Cr-release assay

DES and 2C spleen cells from DES-TCR or 2C-TCR transgenic (TG) mice were depleted of CD4 cells by complement-mediated lysis (GK1.5, anti-CD4 antibody in guinea pig serum 45 min) and 1·5 × 10⁷ cells were in vitro stimulated with 6 × 10⁶ 3000 rad irradiated spleen cells from either C57Bl6 or BALB/c, respectively, in 15 ml RPMI supplemented with 10% FCS, 50 IU/ml penicillin and streptomycin, 0·3 g/l l-glutamine, 1 mm sodium pyruvate, 50 mm 2-mercaptoethanol (2-ME) and 5 units/ml lymphocult-T [interleukin-2 (IL-2) supplement; Biotest Ltd, Solihull, West Midlands, UK].

For the HLA-B27 specific line, two BALB/c female mice were primed by intraperitoneal injection of approximately 3 × 10⁷ B27β2mBALB/c (carries both HLA B27 and human β2m on a BALB/c background) irradiated spleen cells. Three weeks later, bulk cultures were set up as above, but using the HLA B27 spleen, and maintained with fresh stimulators every 7–10 days.

For influenza A virus nucleoprotein (NP)-specific CTL line spleens were obtained 2 weeks after intranasal infection with A/X31 influenza A virus of BALB/c or C57Bl6 mice for in vitro restimulation. Autologous splenocytes were incubated with 1 µM NP147–155 (TYQRTRALV) or NP366–374 (ASNENMETM) (Research Genetics Inc., Huntsville, AL) peptide, respectively, in RPMI at 37° for 1 hr and used as stimulators. In vitro restimulation cultures were set up with 1·5 × 10⁷ splenocytes and 0·3 × 10⁷ peptide-pulsed stimulators in lymphocult-T supplemented medium, as above. The cultures were maintained at 37°, 5% CO₂ for 5 days at which time a standard ⁵¹Cr-release assay was performed. Target cells were labelled with ⁵¹Cr, washed three times in serum-free medium and either infected with A/X31 virus (0·5 ml allantoic fluid for 2 × 10⁶ cells) for 60–90 min or pulsed with 1 µM peptide or unpulsed as indicated. Peptides for 2C recognition in CTL assays were QL9 (QLSPFPFDL)¹³ and SYN (SIYRYYGL)¹⁴ both synthesized in the peptide facility, Institute for Animal Health. Some target cells were incubated in serum containing medium for various times prior to setting up in a standard ⁵¹Cr-release assay.

CTL lines were maintained by re-stimulation every 7–14 days by culturing the effector cells with stimulators, as above, in a ratio of 1 : 2.

Results

Expression of MHC on chondrocytes

Primary chondrocytes isolated from the ventral parts of the ribs of neonatal mice were positively identified by intracellular staining with an anti-collagen type II antibody (Fig. 1a). When chondrocytes were isolated from neonates bred in SPF conditions, there was low or negligible surface staining of MHC class I and class II antigens; however, treatment of the cells with IFN-γ for 48 hr up-regulated the surface expression of both (Fig. 1c, d). We do not believe that the lack of expression is a result of enzymatic effects in the preparation as some cell isolates prepared from mice bred under conventional conditions did show some MHC class I expression ex vivo in the absence of IFN-γ treatment (data not shown).

Figure 1.

Identification and MHC cell surface staining of ex vivo preparations of neonatal rib chondrocytes. (a, b) Intracellular staining for collagen type II on chondrocytes or splenocytes, respectively. Dashed line, polyclonal goat anti-CII antibody (Southern Biotechnology Associates, Cambridge Bioscience, Cambridge, UK); solid line, 2nd step alone, rabbit anti-goat immunoglobulin G fluoroscein isothiocyanate (Sigma). (c, d) Effect of IFN-γ treatment on MHC class I or class II expression, respectively. Bold line, 48 hr IFN-γ treatment prior to staining; dotted line, no IFN-γ, solid line, 2nd step antibody control.

CTL responses to chondrocytes

Initial studies, using virus-specific CTL to look for lysis of peptide pulsed chondrocytes (from SPF mice) in a standard ⁵¹Cr-release assay, showed generally low levels of specific lysis in the absence of pretreatment with cytokines (data not shown). This was not surprising given the low levels of MHC class I expressed on the surface. Following treatment with IFN-γ, peptide-pulsed chondrocyte target cells were efficiently lysed by an influenza A virus NP-specific D^b-restricted CTL line (Fig. 2a). In addition, alloreactive T-cell lines, which were either K^b-specific, derived from the DES-TCR TG mouse, or L^d-specific, derived from the 2C-TCR TG mouse were able specifically to lyse C57Bl6 or BALB/c chondrocytes, respectively (Fig. 2b, c), but not chondrocytes with the self-H-2 type of the CTL. Figure 2(c) also shows recognition by 2C CTL of C57Bl6 chondrocytes pulsed with the syngeneic peptide (SYN), previously identified as a target for high-affinity recognition by 2C on the selecting background.¹⁴ The 2C TCR has also been shown to bind the allogeneic class I MHC L^d in association with an α-ketoglutarate dehydrogenase peptide known as QL9.¹³ Figure 2(c) also shows an increase in lysis of the IFN-γ treated BALB/c chondrocytes when pulsed with this peptide.

Figure 2.

Specific recognition and CTL lysis of IFN-γ treated chondrocytes. (a) A NP366–374-specific CTL line; (B) DES CTL; (C) 2C CTL. CTL were incubated with ⁵¹Cr-labelled chondrocytes, which had been previously treated with IFN-γ for 48 hr and extensively washed. A standard ⁵¹Cr release assay was performed. Chondrocyte targets were, □, unpulsed C57Bl6; ▪, C57Bl6 chondrocytes pulsed with NP366-374 peptide; ○, unpulsed BALB/c; •, BALB/c chondrocytes pulsed with QL9 peptide; ▴, C57Bl6 chondrocytes pulsed with SYN peptide.

Peptides involved in the alloreactive recognition of MHC class I molecules, where identified, are often from abundant self proteins. Thus, it was possible that the low levels of MHC class I on unstimulated chondrocytes could be sufficient for allo-recognition. However we found that DES CTL were only able to recognize chondrocytes from C57Bl6 mice following IFN-γ treatment (Fig. 3a). Similarly, a mouse-derived, xenoreactive cell line specific for HLA B27 was only able to recognize HLA B27 TG chondrocytes following IFN-γ treatment (Fig. 3b). In contrast, pulsing B6 derived chondrocytes with the SYN peptide did allow recognition by the 2C T-cell line (Fig. 3c). This is despite the fact that these same chondrocytes were not recognized by DES CTL (Fig. 3a). Thus, although there were insufficient relevant class I complexes on the surface for DES recognition from endogenously processed peptides, there was sufficient for recognition by 2C when pulsed with the high affinity SYN peptide. In addition, Fig. 2(d) shows some recognition of untreated H-2 D chondrocytes when pulsed with the QL9 peptide. These target cells are the same as for Fig. 2(b).

Figure 3.

IFN-γ pretreatment is required to allow lysis by alloreactive CTL in the absence of added peptide. (a) DES CTL; (b) mouse derived HLA-B27 specific xenoreactive CTL; (c, d) 2C CTL. Chondrocyte target cells, as indicated, were preincubated for 48 hr. with IFN-γ or medium alone. All target cells were labelled with ⁵¹Cr and were further preincubated with 1 µm peptide or medium as indicated prior to setting up a standard 6 hr ⁵¹Cr-release assay. (a) and (c) were simultaneous assays using the same preparations of C57Bl6 and BALB/c derived chondrocytes. (b) and (d) were simultaneous assays using the same preparation of HLA B27/hβ2m BALB/c derived chondrocytes.

CTL are able to recognize influenza A virus infected chondrocytes

We have shown that IFN-γ-treated chondrocytes are recognized by a number of different alloreactive CTL lines without the addition of a cognate peptide. However, these types of interactions may involve the recognition of MHC class I molecules with peptides derived from abundant cellular proteins or with efficient MHC binding properties. In order to show recognition of non-self antigens, chondrocytes were infected with influenza A virus, A/X31. Representative experiments are shown in Fig. 4. In general, the recognition of virus infected chondrocytes was far less efficient than that of peptide pulsed targets, however, the D^b epitope, NP366–379 (Fig. 4a), was consistently better presented than the K^d epitope, NP147–155 (Fig. 4b). Despite the poor levels of lysis of BALB/c chondrocytes, infection of the chondrocytes was confirmed by staining for expression of the NP protein by immunofluorescence, where greater than half the cells showed good levels of expression (data not shown) and were comparable to infection of the standard target cell line P815. Thus, the lack of epitope expression was not due to failure of expression of the influenza A virus NP. In order to make a direct comparison (H-2^d × ^b)F1 progeny were used for preparation of the chondrocytes. Cells were infected with A/X31 virus, and incubated for 0, 120, or 240 min in serum containing medium before the CTL assay was assembled. P815 and EL-4 target cells were similarly treated. EL-4 cells were maximally lysed without a period of incubation following flu infection (Fig. 4c). In contrast, NP147–155-specific lysis of the P815 cells required a period of incubation following infection, and the efficiency of lysis increased with longer incubation times (Fig. 4d). Lysis of the infected chondrocytes by D^b restricted, NP366–379-specific T cells was evident without a period of post-infection incubation, although the efficiency increased with time (Fig. 4e). Using the same chondrocytes, significant levels of lysis by the K^d restricted, NP147–155-specific line were only seen after 240 min of incubation (Fig. 4f).

Figure 4.

Presentation of NP147–155 epitope on K^d by influenza A virus infected chondrocytes is less efficient than presentation of the NP366–374 epitope on D^b. (a) D^b restricted NP366–374-specific CTL line; (b) K^d- restricted NP147–155-specific CTL line. BALB/c or C57Bl6 derived chondrocytes, which had been pretreated with IFNγ, were either uninfected (○), pulsed with either NP147–155 (BALB/c) or NP366–374 (C57Bl6) (•), or infected with influenza A virus A/X31 in serum-free medium (▴). After washing, and prior to setting up the assay, all target cells were incubated for a further 90 min in complete medium containing 10% FCS. (c–f) time course of post infection incubation. (c) and (e) D^b restricted NP366–374-specific CTL line; (d) and (f), K^d- restricted NP147–155-specific CTL line. Target cell lines (as indicated) or chondrocytes derived from (C57Bl6 × BALB/c) F1 mice were uninfected, pulsed with the relevant peptide or infected with A/X31 virus in serum-free medium for 2 hr. Cells were then incubated for 0, 120, or 240 min. before inclusion in standard 4 hr ⁵¹Cr release assays. Peptide pulsed (•) and uninfected (○) are shown as means of the three time points with error bars representing one standard deviation. For virus infected cells:▵, 0 min; ▴, 120 min; ▪, 240 min.

Discussion

In this study, we have used primary cultures of chondrocytes from neonatal mice to determine their susceptibility to recognition and lysis by MHC class I restricted CTL. When chondrocytes were prepared from healthy mice that had been bred under SPF conditions, they had very low levels of MHC class I expressed on their surface. This rendered them relatively resistant to lysis by CTL. However, treatment with IFN-γ caused sufficient up-regulation of MHC to allow recognition and lysis by both alloreactive and virus-specific CTL. One of the alloreactive CTL lines, derived from the 2C-TCR TG mouse, whilst unable to recognize BALB/c chondrocytes unless they were IFN-γ treated, was able to lyse them when pulsed with peptide. This indicates that the low levels of cell surface MHC class I were able to present peptide if sufficient was available and, more importantly, that the untreated chondrocytes were not intrinsically resistant to CTL lysis. Interestingly, some cultures of chondrocytes prepared from neonates bred under conventional conditions did express MHC class I ex vivo, and in this case they were able to present antigens to virus-specific T cells without the need for pretreatment of chondrocytes with IFN-γ (data not shown). This was never seen in preparations from the SPF animals of the same strains. This indicates that the up-regulation seen in vitro is a physiological response that also occurs in vivo. The ability of chondrocytes to process and present both endogenous and viral antigens also confirms them as possible target cells for CTL in the enthesis or joint during localized inflammation or infection.

Cartilage is considered to be an immunologically privileged tissue, with chondrocytes normally encased in an impervious matrix that prevents their interaction with immunoglobulins, lymphocytes and other cells in the synovium. However, under pathological conditions, the integrity of the cartilage is breached and it is gradually destroyed during a chronic inflammatory process. It is thought likely that this process is immunologically mediated¹⁵ and directed against tissue-specific antigens, such as type II collagen and cartilage proteoglycans. The evidence for this is based on the presence of humoral and cellular immunity to these elements in patients^16–19 as well as their arthritogenic potential in animal models of arthritis.^20–23 The chondrocyte is the cell type responsible for the production of such antigens, and it has been suggested that articular chondrocytes may play an active role in the destruction of cartilage in inflammatory joint diseases by presenting antigen to T cells. Indeed, a number of studies have looked for MHC class II restricted responses against chondrocytes. In rabbits, chondrocytes from normal articular cartilage were found to express MHC class II, and could process and present ovalbumin to primed T cells. They were even capable of stimulating proliferation of allogeneic and autologous normal lymphocytes.²⁴ Similar findings were obtained with rat chondrocytes.^25,26 However, variations in chondrocyte phenotype between different species are evident as chondrocytes isolated from humans,^7,8,27 pigs²⁸ and mice²⁹ do not appear constitutively to express MHC class II molecules and require activation with IFN-γ before they can present antigen to CD4⁺ T cells. In contrast to the results obtained with rabbits and rats, human chondrocytes, IFN-γ-treated or not, were very poor inducers of an allogeneic or autologous T-cell response and were able to present only to an antigen-specific T-cell line.^7,8 Studies in pigs have shown that chondrocytes activated by inflammatory cytokines in an arthritic joint in vivo express MHC class II and elevated levels of intercellular adhesion molecule-1 at the cartilage/inflammatory synovium junction.²⁸ The same observation has also been made in rheumatoid cartilage in humans.³⁰

In ankylosing spondylitis and the related spondyloarthropathies, the predominant site of inflammation involves fibrocartilaginous entheses (reviewed in 1). In common with articular cartilage, chondrocytes in this tissue would not normally be in contact with bone marrow or the vasculature. In addition, it is likely that low levels of MHC class I would protect against CTL recognition, should contact occur. However, mechanical damage, stress induced remodelling or pre-existing inflammation could lead to exposure of some chondrocytes to cells from new vessel formation or bone marrow. In an ongoing enthesitis there is an inflammatory infiltrate present and often osteitis or subchondral bone marrow inflammation. Access of CTL to chondrocytes would be possible under these circumstances. The CTL could not only help to further the tissue destruction, but cytokines such as TNF-α and IFN-γ would contribute further to the inflammatory process.