Apoptosis of Fas^high CD4⁺ Synovial T Cells by Borrelia-reactive Fas-ligand^high γδ T Cells in Lyme Arthritis

Lyme arthritis patients came from areas endemic for Lyme disease and were followed at the Lyme Disease Clinic at the University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School. All patients had histories, exams, and serologies consistent with Lyme arthritis, including Borreliaspecific antibody titers that were higher in synovial fluid relative to serum. Synovial fluid lymphocytes were examined from seven patients with Lyme arthritis of 6-mo to 3.2-yr duration.

Lymphocytes were isolated from peripheral blood or synovial fluid by Ficoll–Hypaque centrifugation. Cells were stained with the indicated fluorochrome-conjugated antibody at 4°C for 30 min. Antibodies were specific for TCR-αβ (JOVI-1; Ancell Corp., Bayport, MN), TCR-γδ (5A6.E9; T Cell Sciences, Inc., Cambridge, MA), TCR-Vδ1 and TCR-Vδ2 (AB and BB3, respectively, courtesy of Dr. Alessandro Moretta, University of Genoa, Genoa, Italy), CD4 (SFCI12T4D11; Coulter Corp., Hialeah, FL), CD8 (SFCI21Thy2D3; Coulter Corp.), Fas (M38) (24) and Fas_L (polyclonal C-20; Santa Cruz Biotechnology, Santa Cruz, CA; or monoclonal A11 [25]). Surface staining for Fas_L was performed by one of three methods. The first approach used a fusion protein composed of the extracellular domain of murine Fas linked to the human Ig Fc portion (Fas-Fc) (26). This was followed by goat anti–human Fc–biotin and then avidin–phycoerythrin. Control staining was accomplished by staining for surface IL4 using an IL4 receptor–Fc fusion protein. Alternatively, surface Fas_L was measured using either a rabbit antiserum to the extracellular carboxyl-terminal portion of human Fas_L and purified on a Fas_L sepharose column (C-20), or monoclonal antibody A11 that recognizes both mouse and human Fas (25). To measure Fas_L induction, cells were examined 3 h after stimulation with PMA (10 ng/ml) and ionomycin (250 ng/ml), in the absence or presence of metalloprotease inhibition using 5 mM EDTA (27). Samples were analyzed on a Coulter Elite flow cytometer (Coulter Corp.) and at least 2 × 10⁴ events were accumulated for analysis.

Synovial fluid lymphocytes were cultured in AIM-V serum-free medium (GIBCO BRL, Gaithersburg, MD) in either bulk cultures (10⁶/ml) for phenotyping, or in round-bottomed microtiter wells (10⁵/well) for proliferation assays. Cells were stimulated with 3 μg/ml of a sonicate of B. burgdorferi grown in BSK II medium as previously described (28). Triplicate cultures were pulsed with ³H-TdR during the last 18 h of a 6-d culture, harvested, and counted. From parallel cultures, responding cells were cloned at 0.3 cells/well in AIM-V with 5% FCS in the presence of irradiated PBL (3 × 10⁵/well), human recombinant IL2 (10 U/ml), and 3 μg/ml of B. burgdorferi sonicate. Responding wells were phenotyped and the γδ cells expanded by restimulation at 10-d intervals.

Semi-quantitative PCR was performed on samples using cDNA prepared from oligo-dT–primed RNA and reverse transcriptase (GIBCO BRL) as previously described (29). The 5′ Vδ- and Cδ-specific primers are modifications of published sequences (30) as follows: Vδ1: 5′-AGCAACTTCCCAGCAAAGAG-3′; Vδ2: 5′-AGGAAGACCCAAGGTAACACAA-3′; Vδ3: 5′-CACTGTATATTCAAATCCAGA-3′; Vδ4: 5′-TGACACCAGTGATCCAAGTTA-3′; Vδ5: 5′-CTGTGACTATACTAACAGCATGT-3′; Vδ6: 5′-TATCATGGATTCCCAGCC-3′; 5′Cδ: 5′-CTTGTCTGGTGCAG-3′; 3′Cδ: 5′-CTTCACCAGACAAGCGACAT-3′. A PCR reaction master mix that was common to all samples contained 100 mM Tris HCl, pH 8.3, 500 mM KCl, 2 mM MgCl₂, 200 μM dNTPs, with 25 pmoles of 3′ Cδ primer, 2.2 μCi α-³²P-dCTP, and 2.5 U Taq polymerase (GIBCO BRL) per tube. The final volume was 100 μl and contained 10 ng cDNA, and 25 pmoles of individual Vδ primer. Samples were run on a thermocycler (model 9600; Perkin-Elmer Corp., Norwalk, CT) for 24 cycles using the parameters: cycle 1: 94°C × 3 min, 50°C × 45 s, 72°C × 1 min; cycles 2–23: 94°C × 30 s, 50°C × 45 s, 72°C × 1 min; cycle 24: 94°C × 30 s, 50°C × 45 s, 72°C × 7 min. Samples were resolved on a 29 cm 10% polyacrylamide gel containing 7 M urea in TBE buffer and electrophoresed at 80 V for 18 h. The gel was dried and developed on an analyzer (Betascope 603; Betagen, Waltham, MA). The percentage expression of each Vδ was assigned by dividing the actual cpm for a specific Vδ by the total cpm for Vδ1–Vδ6 after correction for the total Cδ message in each sample.

Fas^low variants of the wild-type Jurkat T cell line, H7 (3% normal surface Fas levels) and B4 (1% normal Fas levels), were derived through irradiation mutagenesis using five doses of 200 Rads each, delivered at 5-d intervals. After each irradiation, cells were cultured in wells coated with lytic anti-Fas antibody (M2, 3 μg/ml)(24). The Fas^low variants and wild-type Jurkat cells were incubated with ⁵¹Chromium (⁵¹Cr) for 1 h, washed, and then mixed at various effector/target ratios with cloned Vδ1 cells in a total volume of 200 μl. After a 4-h incubation at 37°C, 100 μl of supernatant were removed and counted for γ emission. Spontaneous release was determined from labeled targets in the absence of effector cells. Maximum release was determined by lysing target cells with 1.0 N HCl. The percentage of specific ⁵¹Cr release was calculated as:

Blocking studies of cytolysis were performed using either specific antibodies at the concentrations indicated, or Fas-Fc fusion protein (10 μg/ml) preincubated with appropriate cells for 30 min before beginning the cytolysis assay. Antibodies used were specific for TCR-γδ (5A6.E9), HLA class I (W6/32; Accurate Chemical and Science Corp., Westbury, NY), HLA class II (L243; Becton Dickinson & Co., Immunocytometer, Sys., Mountainview, CA), LFA-1 (R7.1; Biosource International, Camarillo, CA), or Fas (M38).

Cells were initially stained for expression of surface γδ, CD4, or CD8 and then fixed for 15 min in 1% paraformaldehyde. Cell membranes were then permeabilized for 15 min using 70% ethanol at 4°C. Samples were incubated at 37°C for 1 h in 100 μl containing 10 U terminal deoxyribosyltransferase and 0.5 nM dUTP-biotin (Boehringer Mannheim Biochemicals Corp., Indianapolis, IN) (31, 32). Specimens were washed twice with PBS/1% BSA and incubated with a 1:50 dilution of avidin-tricolor (Caltag Labs., South San Francisco, CA) at 4°C for 30 min. Cells were washed twice and analyzed by flow cytometry.

Synovial fluid lymphocytes were examined from seven patients with Lyme arthritis of 6 mo to 3.2-y duration. These contained a predominance of CD4⁺ over CD8⁺ αβ T cells in only four of seven cases (Fig. 1,A, Table 1), compared to a consistent CD4 predominance in PBL. Also present in the synovial mononuclear cells was a remarkable percentage of γδ T cells (18.9 ± 6.8%) (Fig. 1,A, Table 1), compared to ∼1–5% in PBL (Reference 1 and see Fig. 3). The synovial γδ population was largely devoid of surface CD4, and only a minor proportion (∼20% on average) expressed low to intermediate levels of CD8 (Fig. 1,B). In addition, whereas γδ T cells from PBL express predominantly the Vδ2 gene product (33), Lyme arthritis synovial fluid γδ cells were primarily of the Vδ1 subset, with lesser proportions of Vδ2 and Vδ3 cells. This was determined by both flow cytometry using Vδ-specific antibodies (Fig. 1,A), and semi-quantitative PCR using specific Vδ primers (Fig. 2).

Stimulation of Lyme arthritis synovial fluid mononuclear cells with a sonicate of B. burgdorferi (strain N40) induced vigorous proliferation (Table 1), yielding a two- to threefold increase in cell number over 6 d. During this period, the composition of T cell subsets shifted considerably. Although the percentage of CD8⁺ cells changed only slightly, there was frequently a striking loss in the proportion of CD4⁺ cells by as much as threefold. Thus, despite the increase in total lymphocyte number during the 6-d culture, there was frequently little change or even a decrease in the absolute number of CD4⁺ cells, as illustrated by patient no. 2 in Table 2. This was paralleled by a reciprocal increase in γδ T cells, in some cases to as much as 50% of the cultured synovial lymphocytes (Fig. 1, Table 1). These continued to be mostly Vδ1 cells as determined by both antibody (Fig. 1,A) and PCR (Fig. 2) analysis.

The loss of CD4⁺ synovial cells might have resulted from unresponsiveness of this subset to B. burgdorferi, and hence overgrowth by the CD8⁺ and γδ⁺ subsets. However, this seems unlikely since we have previously observed that PBL also proliferate strongly to B. burgdorferi with an expansion of predominantly CD4⁺ cells (28). Alternatively, because PBL contain only a small proportion of γδ cells (1), the γδ subset might be responsible for the loss of CD4⁺ cells in Borrelia-activated synovial cultures. Consistent with this notion was the one case (patient no. 6) where the percentage of γδ T cells did not increase following stimulation with B. burgdorferi. In this instance, the proportion of CD4⁺ cells actually increased from 36.5 to 51.4% (Table 1).

To more directly address the possibility that synovial γδ cells might be cytolytic toward the CD4⁺ subset, γδ T cell clones were derived from synovial fluids of two Lyme arthritis patients using a sonicate of B. burgdorferi and irradiated autologous PBL. A panel of 18 Borreliaresponsive γδ clones was established, the majority of which express Vδ1 and lack surface CD4 and CD8. DNA sequencing of the δ chain from seven clones confirmed that they all express Vδ1, but were otherwise each unique and contained varying degrees of N region diversity (Roessner, K., manuscript in preparation).

γδ T cells frequently manifest cytolytic activity toward a broad array of target cells (2, 16). Such a spectrum of cytolysis might occur when a target molecule is widely expressed, as is the case with the apoptosis-inducing molecule, Fas (17). As shown in Fig. 3, Fas expression by fresh CD4⁺ PBL was low to negligible, but was present on a large proportion of CD4⁺ synovial lymphocytes. By contrast, the CD8⁺ and γδ⁺ subsets of PBL or synovial lymphocytes displayed considerably lower levels of Fas.

Surface expression of Fas_L protein by B. burgdorferi–reactive γδ and CD4⁺ αβ T cell clones was examined by flow cytometry using two methods, a Fas-Fc fusion protein as well as a purified anti–human Fas_L rabbit antiserum. Control staining for Fas-Fc was determined using a human IL4 receptor-Fc (IL4R-Fc) fusion protein (as surface-bound IL4 would not be anticipated for a secreted cytokine). Fig. 4,A (column 3) illustrates results of staining using the FasFc fusion protein, on represetative αβ (114B) and γδ (2.11) synovial T cell clones. By this method, surface Fas_L protein was expressed on a considerably higher proportion of the γδ cells than on the B. burgdorferi–reactive αβ T cell clones seven days after the last stimulation. Similar findings were seen with an additional two αβ and two γδ synovial T cell clones. In contrast, the levels of surface Fas antigen on the γδ clones were somewhat less than on the αβ clones, (Fig. 4 A, column 4).

The anti-Fas_L antibody confirmed the disparity in surface Fas_L expression between synovial γδ versus αβ T cell clones. Fig. 4,B (column 1) shows that 7 d after antigenic stimulation of the Borrelia-reactive αβ (114B) and γδ (2.11) clones, surface Fas_L was present on the γδ clone, but was only marginally detectable on the αβ clone. This finding was consistent for three αβ and three γδ clones studied. However, the αβ clones were capable of induction of Fas_L upon stimulation, as shown after 3 h of activation with PMA and ionomycin. In agreement with a recent report (27), Fas_L expression on the T cell line, Molt 4, was enhanced by blocking metalloprotease activity with EDTA (Fig. 4,B, column 4). This was less consistently observed for the αβ T cell clones, and was not observed for the γδ clones. It was particularly striking that the levels of Fas_L on the γδ clones remained detectable for at least 10 d following stimulation with B. burgdorferi (Fig. 4 B, column 1). This is in distinct contrast to αβ T cells which express Fas_L only transiently after activation (21; Roessner, K., unpublished observations).

To further explore whether the Lyme arthritis synovial fluid γδ T cell clones might be cytolytic toward T lymphocytes expressing high levels of surface Fas, the Jurkat T cell line was initially used as a representative Fas^high target. Fig. 5,A shows that the γδ clones manifested very efficient cytolytic activity toward Jurkat cells, with 50% maximal lysis achieved at an effector/target ratio between 10:1 and 3:1. This finding was remarkably consistent for each of five different Vδ1 clones tested from two patients. In contrast, Borrelia-reactive CD4⁺ αβ T cell clones manifested little, if any, cytolysis of Jurkat cells (data not shown). Cytolysis by the γδ clones was not inhibited by antibodies to TCR-γδ, HLA class I or II, but was blocked by anti–LFA-1 antibody (Fig. 5 C ), supporting the notion that cytolysis was dependent on cell contact.

The potential contribution of Fas to cytolysis by γδ cells was examined using three approaches. Initially, two Fas^low variants of Jurkat cells, H7 and B4, were independently derived by radiation mutagenesis followed by selection with lytic anti-Fas antibody, M2. H7 expresses 3% of the levels of Fas found on wild-type Jurkat cells, whereas B4 displays 1% (Fig. 5,B). Fig. 5 A demonstrates that the efficiency of cytolysis of both Fas^low variants was diminished approximately two- to threefold compared to that observed with wildtype Jurkat cells. However, lysis of the Jurkat Fas^low variants was not completely eliminated, suggesting that part of the cytolytic activity of the γδ clones was independent of Fas. This was supported by anti-Fas antibody blocking studies.

Inhibition of Jurkat cell cytolysis by the γδ clones was also achieved using a nonlytic anti-Fas antibody, M38 (24). Fig. 5,D shows that the blocking of cytolysis with M38 was partial, achieving 30–50% inhibition at the highest concentration of antibody (10 μg/ml), whereas control mouse Ig did not block cytolysis. In vitro cytolysis consists of a calcium-independent component mediated by Fas and a calcium-dependent component delivered by perforin (18–20). Blocking perforin action by chelation of calcium with EGTA also resulted in partial inhibition of Jurkat cytolysis, which could then be blocked almost completely by the further addition of anti-Fas antibody (Fig. 5,D). A third method of disrupting Fas-Fas_L interaction used the Fas-Fc fusion protein. Fig. 5 E shows that Fas-Fc, but not IL4RFc, partially blocked cytolysis of Jurkat cells by the γδ clones, though to a slightly lesser extent than did nonlytic anti-Fas antibody.

The above findings show that γδ clones derived from synovial fluid express prolonged and high levels of Fas_L and suggest that γδ cells preferentially lyse Fas^high cells. To directly assess whether uncloned synovial γδ cells function in a similar manner, Fas_L expression was determined on synovial lymphocytes after Borrelia stimulation. As shown in Fig. 6 A, 7 d after activation, Fas_L expression was confined exclusively to a major proportion of the γδ cells. Fas_L was still expressed by at least 50% of the synovial γδ cells for as long as 11 d after Borrelia stimulation.

To further assess the contribution of the γδ cells to the loss of synovial CD4⁺ cells, the γδ subset was depleted by flow cytometric sorting and compared to a nondepleted sample of the same specimen after five days of stimulation with B. burgdorferi. During this period, the γδ cells in the nondepleted synovial sample expanded from 4.3% to 11% (Fig. 6,B). This was accompanied by a decreased proportion of CD4⁺ cells, from 35.6 to 25.3%. In striking contrast, the γδ-depleted population contained only 4% γδ cells after 5 d and manifested a predominance of CD4⁺ cells (40.8%)(Fig. 6,B). In addition, the CD4⁺ cells in the 4-day cultures contained a subpopulation of CD4^low cells which comprised a greater proportion of the total CD4⁺ cells in the γδ-replete than the γδ-depleted specimen (Fig. 6,B, arrow inset). These CD4^low cells represented apoptotic cells, as determined by the TUNEL assay combined with surface staining and analyzed by flow cytometry (Fig. 6 C ). Smaller proportions of apoptotic cells were also observed in the CD8⁺ and γδ⁺ subsets. Observations similar to these have been made with γδ depletion of two additional Lyme synovial fluid specimens, as well as by noting a depletion of CD4⁺ cells when Vδ1 cloned T cells were added to cultures of PBL that have been stimulated with B. burgdorferi (data not shown).

To assess whether the appearance of the apoptotic CD4^low subset in the γδ-replete cultures was in part Fas-mediated, Fas_L was blocked using the Fas-Fc fusion protein. Synovial fluid lymphocytes were stimulated with B. burgdorferi in the presence of either no additives, Fas-Fc, or control mouse IgG. As shown in Fig. 7, the appearance of apoptotic CD4^low cells occurred beginning about five days after Borrelia stimulation. The proportion of this subset increased dramatically thereafter in all cultures except that containing Fas-Fc. The findings support the view that the γδ subset induces apoptosis of synovial CD4⁺ cells at least partly through Fas/Fas_L interactions.

The collective observations suggest an immunoregulatory circuit whereby synovial Vδ1 T cells bearing high levels of Fas_L selectively restrict the expansion of infiltrating inflammatory Fas^high CD4⁺ lymphocytes through cytolysis in a Fas-dependent manner. The findings are in agreement with recent studies showing that Fas_L mRNA expression by T cells is highest in the γδ subset (21). Not only were levels of surface Fas_L high on the Vδ1 clones, they remained elevated for considerably longer periods than similarly activated αβ T cells. This may serve to explain the broad spectrum of cytolytic activity that has frequently been observed for many γδ cells (2, 16). The results parallel other recent descriptions of immunosuppression resulting from constitutive expression of Fas_L by Sertoli cells in the testis (22), and by components of the eye (23).

The current findings may also bear on observations that collagen-induced arthritis in mice (34) and adjuvant arthritis in rats (35) are both more severe following administration of anti-γδ antibody. Collagen-induced arthritis is also more aggressive in mice bearing a genetic deletion of the δ locus (Lefrancois, L., personal communication). Similar results have been observed in a model of orchitis in which γδ depletion accelerated the inflammatory response (36). γδ T cells have also been observed to modulate the functional profile of CD4⁺ cells. In certain instances this has manifested as selectively inhibiting T_H2-dependent cytokine responses, such as IgE production in an allergy model (37) and Coxsackievirus-induced myocarditis (38). The resulting T_H1 bias may be due solely to the production of the T_H1-type cytokine, IFNγ, by γδ cells (37), but may also reflect a greater sensitivity of T_H2 cells to Fas-mediated apoptosis. In this regard, it is noteworthy that B. burgdorferi– reactive CD4⁺ T cells from Lyme arthritis patients express a T_H1 cytokine phenotype (39). Studies are in progress to determine whether a T_H1 enrichment results in the residual CD4⁺ synovial T cells following stimulation with B. burgdorferi.

Lyme arthritis synovial γδ T cells also represent a rare instance where γδ T cell clones obtained from a human infectious disease manifest a proliferative response in the presence of the causative agent. This does not establish that Lyme arthritis synovial γδ cells are responding directly to a Borrelial component. It is entirely possible that B. burgdorferi induces the appearance of surface molecules to which Vδ1 cells respond secondarily. Cutaneous lesions in leprosy also contain γδ T cells that react to the causative agent, Mycobacterium leprae (9). The repertoire of γδ cells that react to mycobacterial products is restricted in both humans and mice (11, 40), and in some instances involves recognition of nonpeptide components such as prenyl pyrophosphates (15, 41). Conceivably, γδ cells in Lyme arthritis may also recognize nonprotein components of B. burgdorferi. On balance, the current findings are consistent with the concept that γδ cells participate in the defense against infectious agents while modulating the immune response through Fas-mediated apoptosis.

We thank Colette Charland (The University of Vermont College of Medicine, Burlington, VT) for assistance with flow cytometry and Roberta Christie (The University of Vermont College of Medicine, Burlington, VT) for preparation of the manuscript.