CD4⁺ and CD8⁺ clonal T cell expansions indicate a role of antigens in ankylosing spondylitis; a study in HLA-B27⁺ monozygotic twins

Abstract

Ankylosing spondylitis (AS) is a complex genetic disease in which both MHC and non-MHC genes determine disease susceptibility. To determine whether the T cell repertoires of individuals with AS show signs of increased stimulation by exogenous antigens, CD4⁺ and CD8⁺ T cell subsets of five monozygotic HLA-B27⁺ twins (two concordant and three discordant for AS) and CD8⁺ T cell repertoires of three healthy HLA-B27⁺ individuals were characterized by TCR β-chain (TCRB) CDR3 size spectratyping. Selected TCRB-CDR3 spectra were further analysed by BJ-segment analysis and TCRB-CDR3 from expanded T cell clones were sequenced. In an analysis of all data (519/598 possible TCRB-CDR3 spectra), AS was associated with increased T cell oligoclonality in both CD8⁺ (P = 0·0001) and CD4⁺ (P = 0·033) T cell subsets. This was also evident when data were compared between individual twins. Nucleotide sequence analysis of expanded CD8⁺ or CD4⁺ T cell clones did not show selection for particular TCRB-CDR3 amino acid sequence motifs but displayed sequence homologies with published sequences from intra-epithelial lymphocytes or synovial T cells from rheumatoid arthritis patients. Together, these results provide support for the hypothesis that responses to T cell-stimulating exogenous or endogenous antigens are involved in the induction and/or maintenance of AS.

INTRODUCTION

In a recent genetic analysis of a large cohort of monozygotic and dizygotic twins with ankylosing spondylitis (AS) [1], heritability of AS was estimated to be at least 97%. These data have confirmed that AS is a complex genetic disease and that both MHC genes (HLA-B27 and others) and non-MHC genes are important determinants of disease susceptibility [1,2].

HLA-B27, which is inherited by more than 90% of AS patients, accounts for only a minority of the overall disease [1]. From these genetic studies, it has been concluded that the contribution of environmental factors to AS is minimal or that environmental triggers for AS would at least be ubiquitous. Experimental data from AS and other spondylarthropathies [3,4] and the finding that HLA-B27 transgenic animals do not develop arthritis when raised under germ-free conditions or when congenitally athymic, however, suggest that bacterial antigens may be critically involved in triggering the disease and, in addition, that AS is a T cell-mediated disease. In a previous study [5], discordances in TCR VB family usage between monozygotic twins and differences in the T cell responses to certain bacterial antigens suggested that the disease pathogenesis was driven by exogenous antigens/pathogens.

In this study we investigated whether TCRB-CDR3 size distribution in AS patients is restricted as a result of immune responses to exogenous stimulating antigens. To control for a genetic bias on the TCR repertoire, we studied CD4⁺ and CD8⁺ T cells from monozygotic twin pairs concordant and discordant for AS. Since the TCR-CDR3 must contact a peptide fragment presented by a genetically defined MHC molecule for the αβ T cell to be activated, TCR-CDR3 length analysis provides a measure to assess the antigenic experience of a particular T cell population and has been a powerful tool to analyse the T cell repertoire under normal and pathologic conditions [6,7]. Gaussian distribution of CDR3 length profiles is typical of polyclonal populations, whereas expansions of specific T cell clones or oligoclonal populations as they result from responses to stimulating antigens cause a perturbation of the Gaussian distribution of CDR3 length profiles. In addition, polymerase chain reaction (PCR) products derived from clonally expanded T cells can be extracted from the gels for subsequent analysis of BJ segment usage and sequencing of their CDR3. This not only proves clonal T cell expansion but also allows determination of whether CDR3 sequences of activated T cell clones share private or public motifs.

PATIENTS and METHODS

Patients

Peripheral blood was obtained from five HLA-B27⁺ monozygotic twin pairs (two concordant for AS and three discordant for AS). To increase the number of healthy individuals, peripheral blood was also obtained from three unrelated HLA-B27⁺ individuals without AS. Patient characteristics are summarized in Table 1. As previously reported [5], all patients responded to an advertisement in the newsletter of the German Ankylosing Spondylitis Society (Bechterew Brief). They were examined by two of the authors (E.M.-H., T.H.). Diagnosis was established according to the modified New York criteria. DNA from peripheral blood samples was isolated by standard procedures and monozygosity determined by DNA fingerprinting analysis using the MZ1.3 multilocus probe. Analysis of HLA-A, -B and -C antigens was performed in all patients by standard microlymphocytotoxicity method. DRB1 DBA typing was performed by nested PCR amplification using sequence-specific primers [8].

Table 1.

Patient characteristics

					HLA-
Patients		Age (years)	AS	Disease duration (years)	A	B	C	DR
Twins, concordant for AS	I CA		+	3
		32			2,2	27,62	w2,w3	4,13
	I CB		+	6
	II CA		+	27
		53			31,32	13,27	w2,w6	7,15
	II CB		+	30
Twins, discordant for AS	III DA		+
		66		44	3,28	7,27	w1,w7	4,15
	III Db		−
	IV DA		+
		54		20	2,26	27,44	w2,w3	11,15
	IV Db		−
	V DA		+
		65		27	1,2	8,27	w2,w7	1,7
	V Db		−
Control individuals	H1	75	−			27
	H2	67	−			27
	H3	22	−			27

D, Discordant; C, concordant; capital letters, individuals with ankylosing spondylitis (AS); H1–H3, healthy unrelated individuals.

Immunomagnetic cell separation

Peripheral blood mononuclear cells (PBMC) were isolated using Ficoll–Hypaque density gradient centrifugation (Seromed Biochrom, Berlin, Germany) and cryoconserved in fetal calf serum (FCS)/10% DMSO. Cells were thawed for analysis and washed. To obtain pure CD4⁺ and CD8⁺ T cell populations for CDR3 length analysis, 25–30 × 10⁶ PBMC were incubated with anti-CD4 or anti-CD8 conjugated magnetic microspheres (MiniMACS; Milteny Biotec, Bergisch Gladbach, Germany) and separated according to the manufacturer's recommendations. Purity was > 95% for CD4⁺ T cells and >93% for CD8⁺ T cells as determined by FACS analysis using a commercial anti-CD4/-CD8 antibody reagent (Dual colour CD4/CD8; Coulter Immunotech, Hamburg, Germany).

RNA extraction and reverse transcription

Total cellular RNA was extracted from 3–5 × 10⁶ cells (PBMC, purified CD4⁺ T cells, purified CD8⁺ T cells) using the RNeasy™ RNA extraction Kit (Qiagen, Hilden, Germany) according to the manufacturer's recommendations. As previously described [9], 1 μg of RNA from each sample was converted to cDNA in a volume of 75 μl reaction mixture containing oligo dT (Sigma, Deisenhofen, Germany), M-MLV reverse transcriptase (Gibco BRL, Eggenstein, Germany), RNasin (Promega, Madison, WI) and dATP, dTTP, dGTP, dCTP (Boehringer, Mannheim, Germany) at 125 μm each in 50 mm Tris–HCl pH 8·3, 75 mm KCl, 3 mm MgCl₂, 10 mm DTT (Gibco).

TCRBV-specific CDR3 length analysis

To ensure that comparable amounts of cDNA were amplified from all samples, we first performed a semiquantitative PCR using a 5′ primer (AGGTCGCTGTGTTTGAGCCATCA) and a 3′ primer (AGCCCGTAGAACTGGACTTGACA) specific for the two alternatively used constant loci TCRBC1 and BC2. cDNA from PBMC, CD4⁺ subset and CD8⁺ subset from all individuals was amplified for 22, 24, 26 and 28 cycles and the amount of PCR product determined by densitometry of 2% agarose gels (Biozym, Hameln, Germany). Differences in amount of cDNA were calculated in the linear range and normalized when necessary. Normalized amounts of cDNA from all samples (PBMC, CD4⁺ subset, CD8⁺ subset) were then amplified using a panel of 26 5′ primers specific for 24 different BV families together with a 3′ primer specific for a sequence shared by the constant loci TCRBC1 and BC2 (sequences from Choi et al. (BV1–20) [10] and Genevee et al. (BV21–24) [11] with modifications as published earlier [9]. For each PCR, 5 μl of diluted cDNA (1:3 with dd water) was added to a reaction tube containing the PCR mix (45 μl). PCR mix consisted of 1 × PCR buffer (Boehringer), 12·5 pmol each of the 5′ variable and 3′ constant primers, 0·1 mm each of dATP, dCTP, dGTP and dTTP (Boehringer) and 1·25 U Taq polymerase (Boehringer). PCR was performed for 35 sequential cycles at 94°C for 20 s (300 s first cycle), 60°C for 45 s, and 72°C for 45 s (600 s last cycle). PCR product (10 μl) was subjected to electrophoresis on a 6% polyacrylamide gel. DNA was visualized by silver staining (Silver Sequence™; Promega) and recorded on positive contact films (Typopaque TR-DO; Promega).

Classification of TCRB-CDR3 spectra

For quantitative analysis of T cell clonality each TCRB-CDR3 spectrum was attributed to one of three characteristic types of TCRB-CDR3 length spectra (Fig. 1): (i) spectra showing a Gaussian distribution of band intensities indicating polyclonal T cell populations, (ii) restricted spectra showing a non-Gaussian distribution of band intensities indicating oligoclonal T cell populations in the absence of clearly dominating bands, and (iii) restricted spectra dominated by one or more expanded bands indicating oligo- or monoclonal T cell expansion. Forerunning bands caused by renaturation of excess clonotypic amplicons typically indicated dominant clonal expansions which allowed determination of the expanded T cell clone by direct sequencing. Classification of CDR3 spectra was repeatedly and blindly carried out and checked by independent colleagues. For analysis of TCRBJ usage of T cells within oligoclonal CDR3 spectra, dominating bands were excised from dried gels and transferred to 200 μl dd water. DNA was eluted and further analysed by BV-BJ-specific PCR using primers specific for the 13 alternatively rearranged BJ segments. BJ primer sequences were GTC GAG TCC CAT CAC CAA AAT GCT GGG (BJ1S5) and CCC CGA AAG TCA GGA CGT TGG CCC (BJ2S6) or as published by Waase et al. [12]. DNA template (5 μl) was reamplified by PCR (40 cycles) and 10 μl of the PCR product were separated on 2·5% agarose gel.

Fig. 1.

Categorization of TCRB-CDR3 spectra. All samples are from twin pair IC. TCRBV9 and TCRBV12 CDR3 spectra show a Gaussian distribution of band intensities indicating polyclonal T cell populations (category i). TCRBV13.1 CDR3 spectra show a non-Gaussian distribution of band intensities indicating oligoclonal T cell populations in the absence of clearly dominating bands in CD8⁺ T cells of twin B (category ii) and a restricted spectrum with one dominant band in CD8⁺ T cells of twin A (category iii). TCRBV8 spectra in peripheral blood mononuclear cells (PBMC) and the CD8⁺ subset of twin B are restricted and dominated by two expanded bands (category iii).

TCRB-CDR3 sequencing

TCRBV-BJ-specific amplicons (40 μl) were purified using DNA binding columns (QIAquick PCR Purification Kit; Qiagen) according to the manufacturer's recommendations. DNA was eluted with 30 μl dd water. A 10-μl aliquot was subjected to the cycle sequencing reaction (Silver Sequence™ DNA Sequencing System; Promega). PCR was performed for 60 cycles according to the manufacturer's recommendations using BV-specific primers as sequencing primers. By this method, TCRBV-BJ-specific amplicons from polyclonal CDR3 spectra do not yield readable sequences due to heterogeneous background; however, expanded T cell clones dominating TCRBV-BJ-specific amplicons from restricted CDR3 spectra can be detected and sequenced.

Sequence data and hydrophobicity analysis

3′ regions of TCRBV segments and consecutive NDN-J-C nucleotide sequence were identified. CDR3 were defined to extend from an amino terminal constant cysteine at TCRBV position 89–92 to a carboxy terminal constant phenylalanine (F) preceding a GXG motif within the BJ element. For determination of CDR3 lengths the N-terminal CAS motif which is conserved in most of the TCRBV germ-line segments and the BJ-encoded constant F were excluded to facilitate comparison with data from the literature. Database searches in EMBL and Genebank sequence libraries using FastA and Advanced BlastP procedures were performed to identify related TCRB rearrangements [13]. Blast procedures were performed with standard parameter settings, expectation value was 100. Hydropathy analysis of amino acid sequences was performed as previously described [14]. In brief, hydropathy values were assigned to each amino acid according to the scaling of Hopp & Woods. CDR3 sequences were adjusted to a constant cysteine (P1) by the definition given above.

Statistical analysis

Data analysis was performed using Fisher's exact test for association. (Dr rer. nat. F. Krummenauer, Institute for Medical Statistics and Documentation, University of Mainz, Germany).

RESULTS

TCRB-CDR3 size spectratyping

TCRB-CDR3 size spectratyping was performed on CD8⁺ and CD4⁺ T cell subsets of five monozygotic (MZ) HLA-B27⁺ twin pairs (two concordant for AS and three discordant for AS) and CD8⁺ T cell subsets of three HLA-B27⁺ control individuals (Table 1). Unseparated PBMC were analysed in parallel and served as internal controls (Fig. 1).

To facilitate individual analyses and to allow comparison between patient groups (Fig. 1), three major categories of TCRBV-CDR3 spectra were distinguished, i.e. (i) spectra showing a Gaussian distribution of band intensities indicating polyclonal T cell populations, (ii) restricted spectra with non-Gaussian distribution of band intensities in the absence of clearly dominating bands, and (iii) restricted spectra dominated by one or more expanded bands.

Increased oligoclonality in CD8⁺ and CD4⁺ T cell subsets of AS patients

Since primers specific for 24 BV families and two additional primers for amplification of TCRBV5 and TCRBV13 subfamilies were used, a maximum of 598 CDR3 spectra was achievable from all samples. Of these, 519 (86·8%) were included in the final analyses. In assessing the frequency of polyclonal versus restricted (type ii or type iii) TCRB-CDR3 spectra, we found that AS was associated with increased T cell oligoclonality in both CD8⁺ (P = 0·0001) and CD4⁺ (P = 0·033) T cell subsets (Table 2). Although the frequency of restricted TCRB-CDR3 spectra was consistently higher in AS patients, restricted CD8⁺ T cell spectra were also frequent in B27⁺ healthy individuals. This was in contrast to the CD4⁺ T cell subsets which harboured significantly fewer restricted spectra than the CD8⁺ subsets. In addition, it was of interest that restricted spectra within the CD4⁺ T cell subset were exclusively detected in individuals with AS.

Table 2.

Increase of oligoclonality in CD4⁺ and CD8⁺ T cell subsets of patients with ankylosing spondylitis (AS)

	CD4+P = 0·033 RR 1·452 95% CI: 1·326–1·590		CD8+P = 0·0001* RR 1·986 95% CI: 1·593–2·432
Type of spectra	AS	Healthy	AS	Healthy
Polyclonal (type i)	146	66	67	105
Restricted	15	0	92	28
Type ii	10	0	50	23
Type iii	5	0	42	5
Total	161	66	159	133

Analysis of polyclonal (type i) spectra versus restricted (types ii and iii) spectra.

Fisher's exact test for association. CI, Confidence interval.

Statistics of the CD8⁺ TCRCDR3 analysis without the three non-related HLA-B27. Control individuals are: P = 0·0001, relative risk (RR): 1·845, 95% CI: 1·622–2·005.

Increased oligoclonality in the CD8⁺ and CD4⁺ T cell subset of AS patients was also evident when the frequencies of non-Gaussian spectra were compared between different twin pairs. Thus, restricted TCRB-CDR3 spectra were concordantly frequent in twin pairs concordant for the disease (IA 61%; IB 74%; IIA 72%; IIB 58% in the CD8⁺ subset and IA 22%; IB 9%; IIA 4%; IIB 12% in the CD4⁺ subset). In twin pairs discordant for the disease, restricted TCRB-CDR3 spectra were more frequent in the affected twins (IIIA 61% versus IIIb 22% for CD8⁺ IIIA 0% versus IIIb 0% for CD4⁺) (IVA 32% versus IVb 23% for CD8⁺ IVA 9% versus IVb 0% for CD4⁺) (VA 46% versus Vb 29% for CD8⁺ VA 8% versus Vb 0% for CD4⁺).

Increased T cell oligoclonality in AS patients is unevenly distributed among TCRBV families

Restricted TCRB-CDR3 spectra in the CD8⁺ subset of AS twins were found in all BV families except BV4 and were most frequent in BV2, BV5.1, BV6, BV7, BV8, BV14, BV16, BV17 and BV20. Compared with the frequency of restricted TCRB-CDR3 spectra within the CD8⁺ subset of HLA-B27⁺ individuals without AS, strong differences were apparent for BV2 (6/7 versus 0/6); BV7 (5/7 versus 0/6); BV8 (6/7 versus 0/6); BV16 (6/7 versus 1/6) and BV20 (6/7 versus 1/6). In individuals without AS, restricted TCRB-CDR3 spectra were most frequent in BV1 (4/6); BV9 (4/6) and BV13.2 (3/6). In the CD4⁺ subset, restricted TCRB-CDR3 spectra were exclusively found in AS patients (BV1, BV3, BV6, BV12, BV13.1, BV13.2, BV15, BV18, BV20, BV23).

Comparison of restricted TCRB-CDR3 spectra between MZ twins

Restricted TCRB-CDR3 spectra were considered of potential pathological relevance when they were confined to either the affected twin in disease-discordant twin pairs, or present in both twins of disease-concordant twin pairs. Such restrictions were identified in the CD8⁺ and CD4⁺ populations but were not specific (i.e. CDR3 restrictions confined to one twin were also present in disease-concordant twin pairs and CDR3 restrictions shared between both twins were also present in disease-discordant twin pairs). It therefore remained unclear whether they were related to disease.

BJ usage and CDR3 sequence of expanded T cell clones

To prove that restricted CDR3 spectra represented oligoclonal populations and to determine the BJ segment usage and CDR3 sequence of expanded T cell clones, a total of 27 expanded bands within the CD8⁺ subset from AS patients were excised from the gels and subjected to BJ segment analysis. cDNA was then directly sequenced. In contrast to subcloning the cDNA before sequencing, this approach selected for dominant (monoclonal) T cell expansions, as only these yield a readable sequence. The frequency of BJ segments amplified from the individual bands varied between 1 and 11. TCRBJ segments from cluster 2 were used more frequently than TCRBJ segments from cluster 1. TCRBJ2S1 and TCRBJ2S7 were most frequent and present in more than 70% of all bands analysed (data not shown). BJ segments from these experiments and from nine expanded bands within the CD4⁺ subset from AS patients were directly sequenced. Proving clonal T cell expansion, individual CDR3 sequences were obtained for 16 T cell clones within the CD8⁺ subset and eight T cell clones within the CD4⁺ subset (Table 3).

Table 3.

TCRB-CDR3 amino acid sequences of clonally expanded T cells from patients with ankylosing spondylitis

Clone no.	Patient (CD8+)	TCRB V	TCRB J		CDR3		CDR3 length
1	I CA	13S2	2S7	VYFC	ASSYWLTXYEQYF	GPGTRLTVT	9
2	I CA	17	2S7	FYLC	ASSITTGAYYEQYF	GPGTRLTVT	10
3	I CB	16	1S2	VYFC	ASSLNTGGYYGYTF	GSGTRLTVV	10
4	I CB	16	2S1	VYFC	ASSQDSTVYNEQFF	GPGTRLTVL	10
5	I CB	8	2S1	VYFC	ASSSVNEQFF	GPGTRLTVL	6
6	IV DA	8	2S5	VYFC	ASGRYQETQYF	GPGTRLLVL	7
7	V DA	2	2S7	FYIC	SARAGTGAVNEQYF	GPGTRLTVT	10
8	V DA	14	2S2	LYFC	ASSPFRIGELFF	GEGSRLTVL	8
9	V DA	14	2S4	LYFC	ASRRTADKNIQYF	GAGTRLSVL	9
10	V DA	14	2S7	LYFC	ASSWTSGYEQYF	GPGTRLTVT	8
11	II CA	8	2S1	VYFC	ASSFKNGDCAFF	GPGTRLTVL	8
12	II CA	8	2S2	VYFC	ASSLYITGELFF	GEGSRLTVL	8
13	II CA	9	2S5	VYFC	ASSLELAGGEETQYF	GPGTRLLVL	11
14	II CB	2	1S3	FYIC	SATDKLRDKVKAGNTIYF	GEGSWLTVV	14
15	II CB	8	1S1	VYFC	ASSRQIMNTEAFF	GQGTRLTVV	9
16	II CB	8	2S3	VYFC	ASSLAGIGDTQYF	GPGTRLTVL	9

	Patient(CD4+)	TCRB V	TCRB J		CDR3		CDR3 length
17	I CA	15	1S4	LYFC	ATSEPTGGSEKLFF	GSGTQLSVL	10
18	I CA	15	2S2	LYFC	ATSDLGGSTGELFF	GEGSRLTVL	10
19	I CA	18	1S5	AYFC	ASSPGGNQPQHF	GDGTRLSIL	8
20	I CA	18	2S3	AYFC	ASSPFRGDTQYF	GPGTRLTVL	8
21	I CA	18	2S4	AYFC	ASSPRPGLEVNIQYF	GAGTRLSVL	11
22	V DA	13S2	2S1	VYFC	ASTFHNEQFF	GPGTRLTVL	6
23	V DA	13S2	2S2	VYFC	ASSDTGELFF	GEGSRLTVL	6
24	V DA	18	2S1	AYFC	ASSPAGVRGTGEQFF	GPGTRLTVL	11

 CDR3 were defined by the constant cysteine of the BV segment and the constant GXG motiv of the BJ segment (Rock et al., J Exp Med 1994; 179:323).

CDR3 sequence analysis of expanded T cell clones

TCRB-CDR3 amino acid sequences of clonally expanded CD8⁺ or CD4⁺ T cells from the same individual or matched BV-specific expansion in different individuals with AS did not show conserved amino acid (aa) sequence motifs. TCR-CDR3 sequences of CD8⁺ T cell clones dominant within the BV8 families of two twins concordant for AS (pair IIC) were also different. Hydrophobicity plots of all 24 TCR-CDR3 sequences failed to detect clear homologies but identified five T cell clones from the CD8⁺ subset (Table 3, T cell clones nos 2, 5, 6, 10, 12) with a hydrophobic aa at CDR3 position 95, which may be involved in antigen recognition [15]. Extensive database searches using the CDR3 sequences given in Table 3 as query sequences revealed different degrees of similarities with published human TCRB-CDR3 sequences. The majority of our queries showed incomplete matches with CDR3 sequences derived from unknown biological background or with CDR3 sequences that were obviously irrelevant in the context of AS. However, three sequences (nos 2, 3 and 18) were related to CDR3 sequences of potential interest. Thus, sequences no. 2 and no. 3 that were expressed by CD8⁺ T cells matched (among others) three sequences derived from distinct intestinal intraepithelial lymphocytes (IEL) [16,17]. Furthermore, a TTG motif used by sequence no. 2 was shared with a T cell clone associated with primary biliary cirrhosis [18]. High similarity (92%) of sequence no. 2 was observed with a CDR3 from a CD25⁺ synovial T cell clone derived from a patient with rheumatoid arthritis (RA) [19]. Also of interest, sequence no. 18, used by a CD4⁺ T cell clone showed significant identity (92%) with a CDR3 sequence from a clonally expanded synovial CD4⁺ T cell from a patient with RA [20].

DISCUSSION

The aim of this study was to investigate whether AS is associated with restricted TCR diversity and to obtain sequence information from strongly expanded T cell clones. To minimize effects of heterogeneic immunogenetic backgrounds, CD4⁺ and CD8⁺ T cell repertoires from monozygotic twin pairs were compared.

T cells respond to cognate conventional antigens by clonal proliferation. Thus, clonally activated and expanded T cells within a pathologic situation may indicate the existence of pathogenetically relevant antigens and may then be used as tools to identify and characterize the crucial epitopes. It is therefore of substantial interest to identify selectively expanded T cell clones in inflammatory diseases, particularly when the triggering antigens are unknown, as is the case in AS and related HLA-B27-associated chronic spondylarthropathies (SpA). Furthermore, as suggested by several immunological studies and by the ‘arthritogenic peptide hypothesis’, T cells are thought to play an important pathogenic role in AS. This theory postulates that the induction of autoreactivity by cross-recognition of foreign (e.g. bacterial) and self-antigens by T cells is the underlying aetiopathogenic process of AS. Targeted deletion or anergization of such pathogenic T cells would thus be an attractive therapeutic rationale.

Data from this twin study show a significant increase of restricted CDR3 spectra with clonally expanded T cells in the CD8⁺ subset of AS patients both individually and collectively. Restricted CDR3 spectra of CD4⁺ T cells (although small in total number) with clonally expanded CD4⁺ T cells were also strikingly associated with AS, since they were completely absent in the non-affected siblings in twin pairs discordant for the disease. Clearly, the increased T cell oligoclonality in concordant twins and affected siblings from twins discordant for AS is independent of the genetic background and provides strong evidence that both CD4⁺ and CD8⁺ T cells from AS patients are exposed to increased antigen stimulation.

The relative importance of the CD4⁺ and CD8⁺ T cell populations in the pathogenesis of AS and other human SpAs has yet to be determined. A major role for CD8⁺ T cells has been suggested by the finding that the classical genetic marker for the SpAs, HLA-B27, encodes a MHC class I restriction molecule presenting antigenic peptides to CD8⁺ T cells. Recent data from transgenic rats for both HLA-B27 and a nonapeptide from influenza nucleoprotein NP383–391 in which the NP product displaces a fraction of the endogenous peptide fraction have supported the concept of an ‘arthritogenic peptide’, since these animals have a significantly reduced prevalence of arthritis compared with HLA-B27 transgenic rats [21]. In the human SpAs, we and others have shown that among synovial fluid and peripheral blood-derived T cells from patients with enterobacteria-induced reactive arthritis (ReA), several bacteria-specific [22,23] or autoreactive [9,22] CD8⁺ T cell clones use HLA-B27 as their restriction molecule. Most of these HLA-B27-restricted clones had rearranged a subset of related TCRBV segments [9]. Repetitive cloning of bacteria-specific or autoreactive T cells with identical TCRB-CDR3, and conserved physicochemical CDR3 properties between different T cell clones suggested that such T cells had undergone clonal expansion in vivo and had been activated by a limited number of antigens [9,14].

However, these and other studies in the human SpAs were limited by the fact that, compared with the total number of T cells at the sites of inflammation or within a whole organism, only a very small selection of T cell clones could be isolated, established and characterized in vitro. In addition, statistical biases due to small sample numbers and biases during the culture procedures cannot be completely ruled out. TCR spectratyping as used here avoids many of these problems, since in vivo preactivated and expanded T cells can reliably be detected and characterized. Titrating experiments using T cell clones with known TCRB chain expression revealed that as few as 1 × 10⁴ monoclonal T cells in a background population of 2 × 10⁶ PBMC (0·5% of the total population) lead to the formation of oligoclonal CDR3 band patterns. Of monoclonal T cells, 2·5% (5 × 10⁴) induce a dominant band typical of CDR3 type iii spectra classified in our study as indicating monoclonal T cell expansion (own unpublished data). In another study, we have also demonstrated that such expanded T cell clones can be isolated in a targeted manner, allowing their subsequent functional characterization [24].

The strong increase of T cell oligoclonality in the peripheral CD8⁺ T cell subset of AS patients compared with their unaffected siblings and with non-related control individuals supports the prediction that CD8⁺ T cells play an important role in AS pathogenesis. It is known that CD8⁺ T cell expansions in diseased and healthy individuals are more common than clonal expansions in the CD4⁺ subset. This is particularly true for elderly people [25] and CD8⁺CD45RO⁺[26] and CD8⁺CD57⁺ T cells [27,28]. In our analysis, the relatively advanced age (> 50 years) of the discordant twin pairs was helpful in terms of assuring that the non-affected twins would not subsequently develop the disease later in life [5]. However, the comparatively high age in the study population might partially explain the relatively high rate of oligoclonality in the CD8⁺ subsets from these healthy individuals (21% of analysed spectra). Even so, oligoclonality was seen approximately three-fold more often in patients with AS (59% of analysed spectra showing non-Gaussian band distribution).

Since HLA-B27 is definitely somehow involved in the pathogenesis of AS, it would not be surprising if this restriction molecule and specific peptides bound to it contribute to the observed T cell expansions. As TCRB-CDR3 sequences from HLA-B27-restricted T cells with variable antigen specificities have previously been reported [9,14,29–32], we performed searches in public sequence databases using the TCRB-CDR3 sequences of expanded CD8⁺ T cells from our AS twins as query sequences. However, this search did not reveal unambiguous sequence or physicochemical motif similarities of our sequences with CDR3 sequences of published HLA-B27-restricted T cells. This was remarkable, since some of our CDR3 sequences used TCRBV 13, 14, or 17, BV families which were previously reported to be preferentially used by HLA-B27-restricted T cells [9,30–31]. The question whether the expanded T cells from our patients are restricted by HLA-B27 therefore remains unresolved. The absence of motif sharing might be related to the fact that all sequences published at present are derived from patients with other forms of HLA-B27-associated SpA, namely reactive arthritides [9,32], or even from healthy individuals [29–31], but not from patients with AS. Moreover, the majority of published patient-derived CDR3 sequences were expressed by synovial but not by peripheral cytotoxic T lymphocytes (CTL), a difference that might be important since synovial and peripheral T cell repertoires may differ substantially [14,33]. Interestingly, some of the CDR3 sequences from CD8⁺ T cell clones showed significant similarities to CDR3 from IEL. This finding may relate to the observation that HLA-B27-associated SpA are commonly associated with inflammation of the intestinal mucosa [34]. CDR3 sequence similarities with IEL-derived CDR3 might thus point toward a common stimulating intestinal antigen. In addition, one of our sequences (no. 2) was closely related to a sequence derived from an expanded synovial T cell clone from a patient with RA. This similarity to intestinal and synovial T cell CDR3 further strengthens the hypothesis that T cells expanded in the peripheral blood of AS patients may be circulating within or originating from the pathogenic gut–joint axis.

The second major result of this study is that in addition to CD8⁺ T cell oligoclonality, CD4⁺ T cell expansions are associated with AS. In contrast to the CD4⁺ T cell CDR3 spectra of AS patients, which were oligoclonal (9·3%) and from which expanded T cell clones could be sequenced, there was a complete lack of oligoclonal CDR3 among the CD4⁺ T cell subsets of unaffected twins. This finding is in concordance with published data from RA patients and healthy individuals [12] and it implies a disease-related increase of antigenic stimulation of CD4⁺ T cells in AS. Just as for the CD8⁺ T cell clones expanded in these patients, the antigen specificity of the expanded CD4⁺ T cell clones is not known. Significant CDR3 sequence similarities between one of these clones (no. 18) and RA-derived CD4⁺ T cells might suggest a role for shared antigens in different types of rheumatic diseases. Although the CD4⁺ subset in AS patients showed less effects of stimulating antigens than the CD8⁺ subset, one could speculate that CD4⁺ T helper cells have a central role by failing to regulate CD8⁺ effector T cells which would then cause tissue damage, or by functioning as effector cells themselves as is suggested from findings in ReA or B27 transgenic rats [35,36]. The high sensitivity of the CDR3 spectratyping method used here substantially extends data on the TCR repertoire obtained from the same twin pairs by BV-specific FACS analysis [5].

In conclusion, the nature of the stimulating antigens that led to the expansion of T cells in the AS patients studied can not be clarified by the existing data. However, the genetic homogeneity of the individuals studied and the increased level of T cell oligoclonality in concordant twins and the affected sibling of discordant twins primarily suggests a role of exogenous, environmental antigens, e.g. from infectious agents or the intestinal flora. In that regard, AS-specific inflammation may be caused by both foreign antigens and self antigens, in that infectious agents induce activation of CD4⁺ or CD8⁺ T cells with cross-reactive properties against articular self-antigens, thus causing ‘bacteria-induced autoimmunity’. Our data do not exclude the possibility of a primarily autoimmune driven T cell response in AS without the need for triggering exogenous antigens. With regard to recently described epitopes of potentially autoantigenic peptide epitopes presented by HLA-B27 [37], it will be of major interest to use these and other candidate peptides to stimulate T cells in vitro and compare the CDR3 spectra of such peptide-specific CD8⁺ T cell lines with the oligoclonal CDR3 sequences described in this ex vivo study.