Mouse strains

C57BL/6 (B6) mice and mouse strains deficient for TLR-2 (129Sv/C57BL/6.TLR-2^-/- [28, 29]) or TLR-4 (C57BL/10ScNCr, homozygous for a null mutation of TLR-4, TLR-4^def [27, 30]) were maintained under pathogen-free conditions in the animal facilities of the Max-Planck-Institut für Immunbiologie, Freiburg, Germany. Male and female mice between 7 and 9 weeks of age were used in all experiments, which were conducted in accordance with the ethical guidelines of the Federation of European Laboratory Animal Science Associations.

Enrichment/purification of cells

Generation of cytotoxic T lymphocyte lines (mixed lymphocyte culture)

Generation of primary alloreactive cytotoxic T lymphocytes (CTLs) and restimulation of these cell lines was performed as described [32]. In brief, for the generation of primary alloreactive CTLs in vitro (primary mixed lymphocyte culture [MLC]), responder splenocytes (one spleen, isolated from B6, TLR-2^-/- or TLR-4^def mice) were co-cultured with irradiated (3000 rad) allogeneic stimulator splenocytes from BALB/c mice (H-2^d, 3/4 spleen) in 40 ml of complete cell culture medium (minimal essential medium [Pan Biotech, Aidenbach, Germany] supplemented with 10% fetal calf serum [Sigma-Aldrich, Taufkirchen, Germany], 100 μg/ml kanamycin [Gibco], 10 μg/ml tylosin [ICN, Eschwege, Germany] and 50 μM 2-mercaptoethanol). CTLs were used on day 6 for cytotoxicity assays and restimulated on day 7. Restimulation for secondary MLC was performed by incubating CTLs derived in vitro (5 × 10⁴/ml) with irradiated BALB/c stimulator cells (2.5 × 10⁶/ml) in complete cell culture medium supplemented with IL-2 (10% of supernatant of rat splenocytes, stimulated with concanavalin A [ConA; Amersham Pharmacia Biotech, Freiburg, Germany] plus 20 mg/ml α-methyl-D-mannopyranoside [Roth]). Cells were used for experiments on day 4 or 5 and restimulated on day 7.

For analysis of the composition of these CTL lines, cells were stained with anti-CD4-FITC (H129.19), anti-CD8a-allophycocyanin (53.6.7), anti-B220-PE (RA3-6B2), anti-NK1.1-PE (PK136), anti-CD19-PE (1D3), anti-CD3ε-biotin (500A2), anti-Thy1.2-biotin (53-2.1) (all purchased from Pharmingen) and anti-F4/80-FITC (Cl:A3-1; Serotec, Eching, Germany).

Functional analysis and proliferation assay of purified T cells or CTL lines

Unselected and purified T cells or CTL lines from MLC were incubated in complete cell culture medium for 72 h (T cells) or 24–48 h (CTLs) in 96-well flat-bottomed plates (Nunc, via Multimed, Kirchheim/Teck, Germany; 4 × 10⁴ cells; 200 μl per well) either coated with rabbit anti-hamster (ha) IgG (Dianova, Hamburg, Germany, 0.5 μg per well) and anti-CD3 (145-2C11; cell culture supernatant purified with Protein A–Sepharose; T cells, 3 ng per well; CTLs, 0.03 or 0.3 ng per well) or with rabbit anti-haIgG alone. The cultures were supplemented or not with recombinant Lip-OspA (strain ZS7, S&K, lot OPA152; GlaxoSmithKline, Rixensart, Belgium), recombinant Met-Asp-Pro (MDP)-OspA (delipidated form, ZS7, S&K, lot 46C33; GlaxoSmithKline; 10 μg/ml maximal concentration of each), human recombinant IL-2 (Sandoz, Basel, Switzerland; 50 U/ml), LPS (S. minnesota, R595; C Galanos, Max-Planck-Institut für Immunbiologie, Freiburg, Germany; 1 μg/ml) or ConA (Amersham Pharmacia Biotech; 5 μg/ml). Anti-TLR-2 mAb (clone mT2.5 [33], at 25, 2.5 or 0.25 μg/ml) or the respective isotype control (mouse IgG; Dianova) was added at various concentrations to cell cultures to analyse their inhibitory potential. For the last 20 h of incubation, 1 μCi of [³H]thymidine (Perkin Elmer, Boston, MA, USA) was added to each well. Incorporation of [³H]thymidine was determined by scintillation counting (cell harvester, Inotech [Dunn Labortechnik, Asbach, Germany]; counting system, TRACE 96 [Dunn Labortechnik]). Means ± SEM for three to six individual wells are given.

Isolation of RNA and analysis by LightCycler^®

Purified T cells from B6 mice (ex vivo, purified by cell sorting for Thy1.2-positive cells), whole splenocytes from TLR-2^-/- mice or alloreactive CTLs from B6 or TLR-2^-/- anti-BALB/c derived from in vitro MLC (purified by cell sorting for CD8⁺ cells) were stimulated for 24 h with phorbol 12-myristate 13-acetate (PMA; Calbiochem, Schwalbach, Germany; 2.5 ng/ml) and ionomycin (Calbiochem; 500 ng/ml) or frozen directly in TriReagent (Sigma, Taufkirchen, Germany) for RNA isolation. RNA was isolated with a modified guanidine thiocyanate/acid phenol method [34] with TriReagent in accordance with the manufacturer's instructions. After treatment with DNAse I (Ambion, Huntingdon, Cambridgeshire, UK), up to 2 μg of RNA was incubated with Random Hexamer primers (Promega, Mannheim, Germany; 1 μM) and Omniscript RT (Qiagen, Hilden, Germany; 4 U).

As a control for plausibility the copy number of mRNA for the low-abundance housekeeping gene TBP (TATA-box binding protein) was also determined and was expected to be between 20 and 40 copies per normal resting cell (data not shown).

Measurement of cytokine secretion

Purified T cells (ex vivo) or CTLs from MLC were cultured in 96-well plates as described above, and supernatants were harvested after 60 h (purified T cells) or 6 h (CTLs), pooled (from six wells per group) and frozen at -20°C until analysed. The concentrations of interferon (IFN)-γ, tumor necrosis factor-α, IL-4 and IL-6 in the supernatants were measured in duplicate with enzyme-linked immunosorbent assay (ELISA) kits from Pharmingen; measurements were performed in accordance with the manufacturer's instructions (IFN-γ, tumor necrosis factor-α and IL-6, cytokine sandwich ELISA; IL-4, OptEIA mouse IL-4 set).

Statistical analysis

Statistical significance was calculated with the two-tailed Student's t-test for comparison of means with unequal variances. P < 0.05 was considered statistically significant.

Recombinant Lip-OspA provides co-stimulatory signals to T cells via TLR-2

To determine a direct co-stimulatory effect of bacterial lipoproteins on T cell proliferation, the preparation of T cell populations of high purity and free from B cells and APCs is critical. Accordingly, T cells were enriched from spleens of B6, TLR-2^-/- and TLR-4^def mice by negative selection via FACS sorting, by using a panel of mAbs against surface markers of B cells (B220), NK cells (NK1.1), dendritic cells (DCs) (CD11c/I-A) and macrophages (Mac-1). Sorted cell populations of B6 and TLR-2^-/- mice contained more than 97% T cells and those of TLR-4^def mice more than 93% T cells (Fig. 1c). The percentages of cells positive for the markers B220, Mac-1, NK1.1 and CD11c/I-A were variable between the three selected T cell populations and ranged between 0% and 0.7%.

Subsequently, the enriched T cell populations were incubated in the presence of plate-bound anti-CD3 mAb, at concentrations known to be insufficient for the induction of proliferation [23], together with either Lip-OspA, its delipidated form MDP-OspA, LPS or recombinant IL-2; anti-haIgG served as negative control. Figure 1 shows one representative experiment (out of three with similar results). The successful depletion of APCs, including B cells and macrophages/DCs, was revealed by comparing proliferative responses of the enriched T cell populations to the various stimuli with those of unselected spleen cells. Unselected spleen cells responded as expected [23]: when incubated on plates coated with anti-haIgG, B6 spleen cells proliferated in the presence of both Lip-OspA and LPS, but not in the presence of MDP-OspA or recombinant IL-2, indicating that most responding cells are B cells (Fig. 1a, upper left panel). As expected, proliferation of unselected TLR-2^-/- and TLR-4^def spleen cells was seen only after incubation with either LPS or Lip-OspA, respectively, under similar conditions.

When unselected spleen cells were incubated in the presence of plate-bound anti-CD3 mAb, all three genotypes responded to recombinant IL-2, indicating the expansion of IL-2-responsive TCR-sensitized T cells, in addition to B cells (Fig. 1a, lower left panel) [23, 24]. In contrast, proliferative responses were not observed in enriched T cell populations of all three genotypes when cells were incubated on plate-bound anti-haIgG, independently of the presence or absence of additional stimuli (Fig. 1a, upper right panel). This finding indicates that the enriched T cell populations were devoid of Lip-OspA and/or LPS-sensitive target cells, particularly B cells, macrophages and DCs. As expected from previous studies [23], all three anti-CD3-stimulated T cell populations proliferated in response to recombinant IL-2. However, after anti-CD3 stimulation only T cells from B6 and TLR-4^def mice, but not those from TLR-2^-/- mice, responded to Lip-OspA. Under these conditions the three cell populations did not proliferate in response to MDP-OspA (Fig. 1a, lower right panel). Most importantly, the three T-cell populations also did not respond to LPS in the presence of anti-CD3, indicating that the T cell populations were devoid of APCs, like macrophages and DCs [23].

Together with the fact that unselected spleen cells from B6 and TLR-2^-/- mice that were sensitized with anti-CD3 were responsive to LPS under similar conditions (Fig. 1a, lower left panel), the data support the notion that co-stimulatory signals provided by LPS to T cells are mediated indirectly, most probably via APCs [23, 24]. The co-stimulatory effect of Lip-OspA for anti-CD3-sensitized T cells from B6 and TLR-4^def mice is dose dependent, whereas T cells from TLR-2^-/- mice are again unresponsive to Lip-OspA at any concentration tested (Fig. 1b).

To determine the functional potential of T cells that were stimulated with plate-bound anti-CD3 mAb in the presence of Lip-OspA, supernatants of the respective cultures from B6 T cells were analysed for cytokine activities. From four cytokines analysed (IFN-γ, tumor necrosis factor, IL-4 and IL-6), only IFN-γ was detectable in the range 0.9–1.4 ng/ml in independent experiments. Levels of IFN-γ production were similar when T cell populations from B6 mice were co-stimulated with anti-CD3 in the presence of Lip-OspA or recombinant IL-2. Cytokine activity was not detectable at all when enriched B6 T cell populations were cultured solely on anti-CD3 mAb or incubated in the presence of either Lip-OspA, MDP-OspA or LPS alone (data not shown). Taken together, these data suggest that TLR-2 functions as a co-stimulatory signal for the maturation of TCR-sensitized T cells.

Recombinant Lip-OspA induces proliferation and IFN-γ secretion in CD8⁺cytolytic T effector cell lines via TLR-2 in the absence of TCR engagement

To determine whether Lip-OspA is also stimulatory for T effector cells, alloreactive (anti-H-2^d) CD8⁺ CTL lines derived in vitro from B6, TLR-2^-/- and TLR-4^def mice were analysed. Figure 2 shows one representative experiment (out of three with similar results). After the third restimulation in vitro, the three CTL lines consisted of more than 99% T cells (Thy-1.2⁺), including 93.1–97.4% CD8⁺ and 0.7–2.0% CD4⁺ T cells (Fig. 2 legend). When incubated on plate-bound control anti-haIgG, proliferative responses of CTL lines from B6 and TLR-4^def mice, but not TLR-2^-/- mice, were significantly increased in the presence of either Lip-OspA or recombinant IL-2 alone (Fig. 2, left panels). When the same CTL populations were seeded on anti-CD3 mAb-coated plates at a concentration insufficient to induce cell growth on its own, again only the addition of Lip-OspA and recombinant IL-2, but not of MDP-OspA or LPS, led to proliferative responses of CTL lines of B6 and TLR-4^def mice but not TLR-2^-/- mice (0.03 ng per well; Fig 2, middle panels). When the CTL lines were stimulated with 0.3 ng of anti-CD3 mAb per well (Fig. 2, right panels) the proliferative responses were increased about 5-10-fold compared with those plated on anti-haIgG or on 0.03 ng of anti-CD3 per well. The three CTL populations did not show proliferative responses to MDP-OspA, ConA or LPS above background, independently of whether they were cultivated on plate-bound anti-haIgG or anti-CD3 mAb. Note that the proliferative response of TLR-2^-/- CTLs with anti-CD3 mAb alone was higher than that of B6 and TLR-4^def CTLs, but that it was not altered by the addition of Lip-OspA.

In general, T cells from TLR-2^-/- mice were more reactive to appropriate stimuli (compare the stimulation indices with recombinant IL-2 in the left and middle panels of Fig. 2) but also seemed to function at an accelerated pace (compare absolute counts in the right panels of Fig. 2). When a mAb against mouse TLR-2 with inhibitory potential [33] was included in the cell culture, a significant and dose-dependent decrease in the proliferative response of anti-CD3 (both 0.3 and 0.03 ng per well) plus Lip-OspA-stimulated B6 CTL lines compared with control cultures (without anti-TLR-2 mAb or in the presence of the isotype control antibody) was observed (Fig. 3).

Quantitative analysis of TLR expression on resting and activated T cell populations

These data demonstrate that naive resting and effector T cells express TLRs appropriate for binding pathogen-associated molecular patterns of B. burgdorferi and are fully compatible with the results shown in Figs 1, 2 and 3.

Specificity of the interaction

The present finding that Lip-OspA directly co-stimulates anti-CD3-sensitized T cells from B6 and TLR-4^def mice, but not from TLR-2^-/- mice, to proliferate and to develop into effector cells explains our previous findings that Lip-OspA augments proliferative and cytokine responses of mouse and human T cells [23, 24]. Here we describe a direct involvement of TLR-2 expressed on T cells as the underlying molecular mechanism. This conclusion is supported by the fact that naive and presensitized T cells are shown to express the respective receptor, although at low levels, in line with previous reports on TLR expression in murine T cell lines [44] and in thymic and splenic T cells [9]. We found that after polyclonal activation with PMA and ionomycin, the expression of TLR-2 increased in CTLs but not in freshly isolated splenic T cells. Whereas the expression of TLR-4 transcripts was not seen in naive T cells, TLR-4 mRNA could be detected after stimulation and was even higher in CTLs.

These findings extend reported data [9] in which an increase in TLR-2 but not in TLR-4 transcripts was observed after stimulation of splenic and thymic T cells. The expression of TLRs that have been described as partners in a heterodimeric complex with TLR-2, namely TLR-1 and TLR-6 [36–38, 41], are also regulated rather markedly: a more than 10-fold decrease in TLR-1 was found in PMA/ionomycin-treated freshly isolated B6 cells, as well as in CTL lines. Whether this TLR modulation would translate into a change of susceptibility for activation by these specific ligands has not been studied.

However, the differential effect of Lip-OspA on B6, TLR-4^def and TLR-2^-/- T cell populations suggests the surface expression of the respective lipoprotein receptor on presensitized T cells and CD8⁺ T effector cells. This assumption is furthered by the fact that co-stimulation of B6 CTL lines by Lip-OspA was inhibited by a mAb against TLR-2, known to interfere with ligand–receptor interaction [33]. It is not yet clear which level of (protein) expression of TLRs in general is necessary for efficient signaling of the target cell, but all evidence points to a low expression of most TLRs [45]; however, this does not seem to interfere with an efficient biological response to a stimulus. In this regard it is significant that TLR-4, which is the receptor for LPS, a polyclonal B-cell activator and inducer of differentiation, is expressed at comparable levels in B cells (Table 3) [25, 26].

Neither naive splenic T cells nor CTL lines responded to LPS. This finding is remarkable for two reasons: first, in view of the fact that the co-stimulatory activity of LPS for T cells is strictly dependent on APCs [23] it verifies the successful enrichment of the responder populations; second, it conflicts with the (albeit low) expression of TLR-4 on CTLs and the reported responsiveness of TLR-4-positive regulatory T cells to LPS [46]. Optimal signaling by LPS requires, besides TLR-4, several accessory molecules such as LBP, MD-2 and CD14 [47, 48]. We do not yet know whether the absence of a response to LPS by T lymphocytes that do express TLR-4 is due to qualitative or quantitative aspects of signal transduction by TLR-4 on T lymphocytes.

Biological effects

The biological effects of TLR engagement on target cells are poorly understood, which is partly due to insufficient knowledge about the signal transduction pathways. Recent evidence indicates that subgroups of TLRs use specific signaling pathways [49–51] aside from common pathways employed by all TLRs, IL-1 receptors and other surface receptors. Evidently CD28, the prototype co-stimulator of T lymphocytes, shares certain signaling pathways with TLRs [52]. Thus, the distinct outcome of a T cell response, such as differential cytokine production, as seen with human T cells co-stimulated by either Lip-OspA or CD28 [24], can be understood by implicating non-overlapping parts of distinct signaling pathways. Additional levels of sophistication seem to derive from a differential expression of TLRs on different T effector cells ([46], and this study) and the dependence of recognition on the physical state of the pathogen-associated molecular patterns. For example, the recognition of OspA by TLR-2/TLR-6 or TLR-2/TLR-1 heterodimers depends on the acylation state of the lipoprotein [36–41, 53, 54]. In addition, little is known about feedback regulation after engagement of TLRs and the consequences of the absence of particular TLRs on effector cells. The increased excitability of TLR-2^-/- CTL lines should be considered in this context.

Current concepts of T cell activation imply that co-stimulatory molecules are necessary to initiate antigen-specific responses from naive T cells but are dispensable for triggering functions in effector T cells, including exocytosis and cytokine release [55]. The present finding that bacterial lipoproteins directly stimulate alloreactive CTLs to proliferate and to secrete IFN-γ via TLR-2 adds another facet to the functional potential of T effector cells. The fact that the same stimulatory signal leads neither to the release of cytotoxic effector molecules from CTLs, such as perforin and granzymes, nor to an enhancement of their cytotoxic potential in the presence of appropriate target cells or anti-CD3 mAb indicates that TCR-induced granula exocytosis is independent of TLR-2 signaling. These findings not only emphasize the differential signal requirements for the induction effector function in T cells [56], including CTLs, such as granula exocytosis and cytokine release [57]; they will certainly also contribute to a better understanding of T cell-driven pathological processes in inflamed tissues, even in situations where causal agents are elusive.

Relevance of the findings

The question of whether our findings are of any significance for the understanding of CIJD is justified and needs answering.

1. Involving microbial infections as a leading cause for CIJD would reconcile years of research in this area and numerous hypotheses on its pathogenesis [4, 14, 58, 59].

2. Recent research implicates synovial lining cells, B cells and T cells in the pathogenesis of CIJD (for a recent review on this, see [2]).

3. The receptor system, implied by our findings, is present on synovial lining cells, B cells and T cells as is shown by our own data and published results [8, 10].

4. TLR ligands have long been known as polyclonal activators of lymphocytes, in particular of B cells [27, 60, 61].

5. TLR ligands have been implicated by other groups as a cause of CIJD or as enhancing factors in the disease, for example hypomethylated bacterial DNA [12], LPS [13] and heat shock protein 60 [62].

6. TLR ligands are found in the synovia of patients with CIJD [11].

7. The cytokine profile in the serum of patients with inflammatory joint disease or produced by T cells isolated from synovia is congruent with that produced by the T cells in our experiments [63–66].

8. In our hypothesis a specific antigen is not required, leaving room for a multicausal hypothesis on the pathogenesis, including T cells of any specificity.

9. The pathogenesis and (histo)pathological findings in B. burgdorferi infection are compatible with those of CIJD [14, 67–70].

These data suggest that the described inflammatory processes are elicited and maintained by direct interaction of intact spirochetes and/or extracellular membrane-bound vesicles [71] and molecules thereof with cells that either home to the affected tissue or infiltrate the diseased area. The observation that susceptibility to the development of chronic arthritis in patients with B.burgdorferi infection is linked, at least partly, to HLA-DRB1*0401 or related alleles [72, 73], just as the predisposition of normal mouse strains with certain H-2 haplotypes to develop chronic joint inflammation [69, 74] indicated the critical involvement of T cells in the pathogenesis of Lyme disease. The fact that human T cells with specificity for a particular OspA epitope in the context of HLA-DR4 protein co-recognize an epitope on a host adhesion molecule, LFA-1, led to the hypothesis that Lyme arthritis could be a consequence of a specific pathogen-induced autoimmunity [75]. However, at present, there is no convincing experimental evidence whatsoever for such a causal relationship [18, 76–78]. In addition, no correlation was found between responses of T cells to LFA-1 peptide in patients with Lyme disease and their clinical status [79].

The findings that synovial T cells from patients with Lyme arthritis are polyclonal [14, 15] and that pre-activated T cells, irrespective of their antigen specificity, effectively infiltrate inflammatory foci [80, 81] suggest that T cells specific both for spirochetal and for third-party antigen can expand and secrete pro-inflammatory cytokines in infected tissue, thereby contributing to the disease progress. Engagement of TLR-2 and other TLRs with resident spirochetes or their products would give any pre-activated T cell a nonspecific stimulus that ensures the ongoing inflammation in a seemingly specific way (Fig. 4). This might also hold true for other pathogen-induced or non-pathogen-induced CIJD.