Autospecific γδ thymocytes that escape negative selection find sanctuary in the intestine

Abstract

αβ or γδ thymocytes whose T-cell receptors (TCRs) recognize endogenously expressed antigens (Ag) are autospecific and, thus, potentially self-reactive. In the thymus, such T cells are eliminated during T-cell development through a process known as negative selection. As a model of negative selection of γδ T cells, we have used G8 γδ–T cell transgenic mice, which express a γδ TCR that recognizes the nonpolymorphic MHC class I TL^b molecule. Here, we demonstrate that negative selection of autospecific γδ T cells is almost complete in the adult thymus but is markedly attenuated in the neonatal thymus. A consequence of this attenuated negative selection is that potentially self-reactive γδ thymocytes are allowed to escape negative selection, undergo extrathymic differentiation, and find sanctuary in the intestinal epithelium. Interestingly, the ability of these potentially self-reactive γδ T cells to find sanctuary requires both the intestinal epithelial environment and the extrathymic presence of the self-Ag. The implications of these findings on the development and persistence of autoreactive T cells in autoimmune disease are discussed.

Introduction

Immature T cells that express an autospecific T-cell receptor (TCR) that recognizes an endogenously expressed antigen are eliminated in the thymus through the process of negative selection. This process has been shown in several systems to be an efficient method by which the body prevents potentially harmful autoreactive T cells from developing. During T-cell development, thymocytes that express a specific Vβ TCR that recognizes an endogenously expressed Mls superantigen are eliminated, and are thus virtually absent from the adult thymus (1, 2). Similarly, studies on transgenic (Tg) mice have shown that mature T cells that express a TCR that recognizes an endogenously expressed antigen (Ag) are also deleted (3–7).

Several lines of evidence suggest that, in contrast with the adult thymus, negative selection of autoreactive αβ and γδ T cells is either absent or markedly attenuated early in ontogeny. T cells that express Vβ TCRs that recognize an endogenously expressed Mls Ag are found in relatively large numbers in the neonatal thymus, but are virtually absent in the adult thymus (8–10). Similarly, in HY, αβ–T cell Tg mice, Teh et al. demonstrated that Tg αβ T cells that recognize the male HY antigen are found in almost equal numbers in the thymus of male and female mice at fetal day 16, but these cells begin to decrease markedly in the male thymus at or near the time of birth (11). Using G8 Tg γδ-TCR mice, Dent et al. demonstrated that Tg γδ T cells, which recognize an endogenously expressed TL^b antigen (T10^b or T22^b), are found in the neonatal spleen, but are virtually absent in the adult spleen and thymus of TL^b+ mice (12). Although their results suggest that negative selection of Tg γδ T cells is either absent or attenuated in the neonatal thymus, Dent et al. did not examine the neonatal thymus and therefore did not address the possibility that Tg γδ T cells found in the spleen of neonatal TL^b+ mice were derived from an extrathymic pathway (13–16). Nevertheless, these results collectively suggest that the ability of the neonatal thymus to remove autospecific neonatal thymocytes is either absent or severely attenuated.

The fate of these autospecific neonatal thymocytes is an important issue that has not been addressed in the discussed neonatal studies. Presumably, a few cells escape out into the periphery and remain senescent at very low numbers, but most die from negative selection and possibly peripheral deletion. Therefore, it is interesting to note that like the neonatal thymus, the intestinal epithelial lymphocyte (IEL) population appears to be rich in T cells that bear autospecific TCR. The IEL population has been shown to be rich in αβ T cells, which express Vβ TCRs that recognize an endogenously expressed Mls Ag (17). Furthermore, studies on αβ– and γδ–T cell Tg mice, which express a TCR that recognizes endogenously expressed Ag, have shown that the IEL population is rich in T cells that express autospecific TCR (16, 18–20). However, how these IEL that express autospecific TCR escape negative selection is unclear. Present theory suggests that most IEL that express an autospecific TCR avoid thymic negative selection by developing by an extrathymic pathway (16, 17, 21). However, the relative contribution of the extrathymic pathway to the development of most IEL is controversial (22, 23). Nude mice that are congenitally athymic have virtually no TCR αβ IEL and have markedly reduced numbers of TCR γδ IEL compared with normal euthymic mice (22, 24). Furthermore, several studies have shown that many IEL, which were originally believed to develop solely through an extrathymic pathway, can also be derived from the fetal/neonatal thymus (25, 26).

In this study, we investigated whether there is a direct relationship between the absence or attenuation of negative selection in the neonatal thymus and the abundant number of IEL that express an autospecific TCR. Using G8 γδ Tg mice as a model of negative selection of γδ T cells, we demonstrate that negative selection is severely attenuated in the neonatal thymus so that up to 40% of day 2 neonatal thymocytes bear the autospecific Tg γδ TCR. Furthermore, a consequence of this attenuated negative selection is that potentially self-reactive Tg γδ neonatal thymocytes, which appear to be in the process of undergoing negative selection, are able to escape, differentiate extrathymically, and find sanctuary in intestinal epithelium, where they survive and incease in number. Lastly, we demonstrate that the ability of the intestinal epithelium to serve efficiently as a sanctuary for autospecific γδ T cells requires, paradoxically, the extrathymic presence of the self-Ag. The implications of these findings on the development of IEL and the development and persistence of autoreactive T cells found in autoimmune disease are discussed.

Methods

Mice.

Transgenic mice with or without the TL^b Ag (Tg^b/d or Tg^d/d, respectively) were generated by breeding a single G8 Tg^d/d founder male to either C57BL/6 (TL^b+, H-2^b+) or BALB/c (TL^d+, H-2^d+) females, respectively (3, 16, 27). Time of birth was considered day 0. All mice, including athymic mice bearing the TL^b Ag (TL^b+, C57BL/6 nu/nu) and those not bearing the TL^b Ag (TL^d, BALB/c nu/nu), were obtained from The Jackson Laboratory (Bar Harbor, Maine, USA). All mice were raised under specific pathogen–free conditions in the animal care facility at the La Jolla Institute of Allergy and Immunology.

Cell isolation, flow cytometry analysis, and cell sorting.

Isolation of IEL and cells from the thymus, spleen, and lymph node has been described previously (28). IEL described specifically as Tg^b/d neonatal thymus–derived were isolated from C57BL/6 nude mice grafted 6 weeks previously with Tg^b/d day 2 neonatal thymus (described below). Two- or 3-color flow cytometry analysis was performed with a FACScan flow cytometer from Becton Dickinson and Co. (Franklin Lakes, New Jersey, USA). The data were analyzed with the Macintosh CellQuest program. Cell sorting was performed with a FACStar cell sorter from Becton Dickinson and Co.

All antibodies were obtained from PharMingen (San Diego, California, USA) unless otherwise noted. Antibodies and reagents used were as follows: FITC-conjugated and nonconjugated anti-Vγ2 (UC3-10A6), FITC-conjugated and biotin-conjugated anti-TCRβ (H57-597), PE-conjugated anti–Thy 1.2, biotin-conjugated anti-CD8β (Caltag Laboratories Inc., Burlingame, California, USA), biotin-conjugated anti–H-2Kd, biotin-conjugated anti–H-2K^b, PE-conjugated anti-CD4, PE-conjugated anti-CD8α, PE-conjugated CD45RB, PE-conjugated anti-HSA, PE-conjugated B220, and streptavidin-PE (GIBCO BRL, Gaithersburg, Maryland, USA).

Neonatal thymectomy and thymus grafting of nude mice.

Neonatal thymectomy was performed on day 2 neonates as described previously (29, 30). Grafting of nude mice was performed by placing either 2 lobes of day 2 neonatal thymus or a portion of adult thymus (trimmed to be of similar size to a pair of day 2 neonatal thymuses) under the kidney capsule as described previously (31).

Proliferation assay.

Unless otherwise stated, responder cells (2 × 10⁵) were cultured with either plate-bound mAbs (3 μg/mL) or stimulator cells (2 × 10⁵ irradiated spleen cells) for 36 hours in 100 μL RPMI with 10% FCS, and then pulsed with 1 μCi of hydrogen-3 for 12 hours before harvest. Assays were performed at day 1, day 2, and day 3, in triplicate.

Cytotoxic assay.

The ability of IEL to induce the cytotoxic killing of human Jurkat T cells (primarily through a Fas-mediated mechanism) has been previously described (32). In brief, Jurkat target cells (10⁶ cells/mL) were labeled with 5 mCi/mL [³H]thymidine for 2 hours. Unincorporated [³H]thymidine was removed by 2 washes with HBSS. IEL effector cells were cultured with labeled Jurkat target cells (2 × 10⁴) at various effector target ratios in flat-bottomed, 96-well plates that had been previously coated with 3 μg/mL of anti-Vγ2 mAb. After 12 hours, cells were harvested using a Skatron cell harvester, and the quantity of [³H]thymidine-labeled unfragmented DNA was calculated as follows: % DNA fragmentation = 100[1 – (cpm experimental group/cpm control group)] ± SD. Assays were done in triplicate.

Measurement of IL-2 from Tg^b/d neonatal thymus–derived Vγ 2 IEL.

IL-2 was quantified from the supernatant of 2 × 10⁵ IEL cultured with stimulator cells (2 × 10⁵ irradiated TL^b+ spleen cells) for 48 hours in 200 μL of RPMI and 10% FCS media, using a murine IL-2 ELISA kit (Endogen Inc., Woburn, Massachusetts, USA), following the manufacturer’s instructions.

Results

γδ T cells expressing autospecific TCR are deleted in the adult thymus but are found abundantly in the IEL population.

In G8 γδ–TCR Tg mice, Tg γδ T cells express the Vγ2⁺ TCR that recognizes the T10^b or T22^b gene product of the nonclassical MHC class I TL region (33, 34). Thus, in Tg mice with an H-2^d+ background (Tg^d/d), neither T10^b nor T22^b is present (TL^b–), and thus Vγ2⁺ T cells are found in very high numbers in the thymus, spleen, and lymph node (Figure 1a and data not shown). However, in Tg mice with an H-2^b+ background (Tg^b/d), T10^b or T22^b is present (TL^b+); in the adult mice, Vγ2⁺ T cells are deleted and thus virtually absent in the thymus, spleen, and lymph node (Figure 1a and data not shown). The absence of Vγ2⁺ T cells in the spleen and lymph node of Tg^b/d mice reflects either efficient negative selection of Vγ2⁺ thymocytes, or (possibly) peripheral deletion of the few Vγ2⁺ thymocytes that escape negative selection in the Tg^b/d thymus. In contrast, γδ T cells found in the IEL population appear to be an exception (27). The virtual absence of Vγ2⁺ T cells in the thymus and periphery of Tg^b/d mice contrasts profoundly with the abundant number of Vγ2⁺ T cells found in the IEL population in both Tg^d/d and Tg^b/d mice (Figure 1b). To explain this paradox, present theory suggests that most Tg^b/d Vγ2⁺ IEL are derived from an extrathymic pathway (16).

Figure 1.

(a) Flow cytometry (FCM) 2-color analysis of thymus and spleen from Tg^b/d (top row) and Tg^d/d (bottom row) mice. (b) FCM analysis of IEL from Tg^d/d (left) and Tg^b/d (right) mice. (c) FCM histogram analysis of IEL from 6-week-old Tg^b/d mice that were sham operated (left) and thymectomized (NTX; right) on day 2 of life. Data shown represent 1 of 2 independent experiments.

The majority of Tg^b/d Vγ2⁺ IEL are thymus dependent.

Instead of assuming that Tg^b/d Vγ2⁺ IEL are derived from an extrathymic pathway, we entertained the possibility that the development of Tg^b/d Vγ2⁺ IEL was dependent on the thymus. Therefore, we thymectomized Tg^b/d mice on neonatal day 2 and examined the IEL population after 6 weeks. Figure 1c demonstrates that neonatal thymectomy resulted in a nearly complete depletion of Vγ2⁺ IEL in adult Tg^b/d mice, suggesting that development of a vast majority of Vγ2⁺ IEL in Tg^b/d mice is dependent on the thymus. In addition, examination of the IEL phenotype of neonatal thymectomy Tg^b/d mice revealed that there were also very few TCR αβ IEL (less than 5%, data not shown). Overall, these results are consistent with our previous observation that neonatal thymectomy results in the depletion of most TCR αβ IEL and TCR γδ IEL (22).

Vγ2⁺ thymocytes are increased in the neonatal Tg^b/d thymus.

Neonatal thymectomy results suggest that most Tg^b/d Vγ2⁺ IEL are derived from the thymus. However, it is unlikely that the adult Tg^b/d thymus is a major source of Tg^b/d Vγ2⁺ IEL, because Vγ2⁺ thymocytes are virtually absent in this organ (Figure 1a). Therefore we examined the Tg^b/d neonatal thymus as a potential source of Tg^b/d Vγ2⁺ IEL. Unlike Tg^b/d adult thymocytes, a large percentage of Tg^b/d neonatal thymocytes express the Vγ2⁺ TCR (Figure 2a). The percentage of Tg^b/d Vγ2⁺ thymocytes varied considerably among the thymuses (ranging between 7% and 40%, data not shown), and persisted at increased levels up to 7 days of age. By 2 weeks of age, the percentage of Vγ2⁺ thymocytes became severely diminished and similar to the level found in adult Tg^b/d thymus (data not shown).

Figure 2.

Negative selection is absent or attenuated in the neonatal thymus. (a) FCM histogram analysis for Vγ2 expression in adult (top row) and neonatal (day 2, bottom row) thymus from Tg^b/d (left) and Tg^d/d (right) mice. Data shown represent 1 of 3 independent experiments. (b) Tg^b/d Vγ2⁺ but not Tg^b/d Vγ2^– neonatal thymocytes can give rise to Vγ2⁺ IEL. Flow cytometry–sorted Tg^b/d Vγ2⁺ (3 × 10⁵) and Vγ2^– (10⁶) neonatal thymocytes were injected into TL^b+ athymic C57BL/6 mice. Six weeks later, IEL of injected athymic mice were examined for the presence of graft-derived (H-2^d+) Tg Vγ2⁺ IEL. Data shown represent 1 of 3 independent experiments.

Tg^b/d Vγ2⁺ neonatal thymocytes can give rise to Tg^b/d Vγ2⁺ IEL.

Some compelling evidence suggests that the murine intestinal epithelium is a site for extrathymic T-cell development (13, 19, 21, 23, 35). Hence, Tg^b/d Vγ2⁺ neonatal thymocytes may not migrate to the intestinal epithelium, but Tg^b/d Vγ2^– neonatal thymocytes could migrate to the intestinal epithelium and develop extrathymically into Tg^b/d Vγ2⁺ IEL. To address this question, flow cytometry–purified Tg^b/d Vγ2⁺ or Tg^b/d Vγ2^– thymocytes were injected into congenitally athymic TL^b+ nude C57BL/6 hosts. Figure 2b demonstrates that Tg^b/d Vγ2⁺, but not Tg^b/d Vγ2^– neonatal thymocytes, are capable of generating Vγ2⁺ IEL. This suggests that it is unlikely that Tg^b/d Vγ2^– neonatal thymocytes migrate to the intestinal epithelium and differentiate into Vγ2⁺ IEL. Furthermore, examination of lymph node and spleen of TL^b+ nude mice injected with Tg^b/d Vγ2⁺ neonatal thymocytes revealed no detectable presence of Tg Vγ2⁺ T cells (data not shown). This suggests that Tg^b/d Vγ2⁺ neonatal thymocytes are predestined to either migrate to the intestinal epithelium or undergo peripheral deletion in the spleen and lymph node.

Efficient migration of Tg^b/d Vγ2⁺ neonatal thymocytes to the intestinal epithelium requires the extrathymic presence of the TL^b Ag.

To explore the fate of Tg^b/d Vγ2⁺ neonatal thymocytes in a model that mimics normal ontogeny as closely as possible, we grafted Tg^b/d (H-2^b+/d+) neonatal thymus into TL^b+ (H-2^b+) adult nude mice. In this model, cells derived from the thymus graft (donor) can be distinguished from cells derived from the nude hosts by a mAb that specifically recognizes the donor-specific class I H-2K^d molecule (shown as donor H-2⁺ in Figure 3). As shown in Figure 3 (upper row), grafting of the Tg^b/d neonatal thymus into a TL^b+ adult nude host readily generated graft-derived cells in both the spleen and IEL populations. Not surprisingly, the Tg^b/d neonatal thymus failed to generate Vγ2⁺ T cells in the spleen of the nude TL^b+ host. Presumably, this is either because Tg^b/d Vγ2⁺ neonatal thymocytes cannot escape negative selection or they undergo deletion after escaping into the periphery. In contrast, the Tg^b/d neonatal thymus was very efficient at generating Vγ2⁺ IEL when grafted into TL^b+ nude hosts. Phenotypic studies of the Vγ2⁺ IEL generated by the Tg^b/d neonatal thymus graft in TL^b+ nude hosts suggests that they are mature (HSA^–) and bear a phenotype that resembles the Vγ2⁺ IEL found normally in Tg^b/d mice (CD8α^d+, Thy 1^–, and B220⁺), rather than the Vγ2⁺ IEL found in Tg^d/d mice (CD8^–, Thy 1⁺, B220^–) (16 and data not shown). In addition, the functional properties of Tg^b/d neonatal thymus graft–derived Vγ2⁺ IEL appear to be very similar to those reported for Vγ2⁺ IEL found in Tg^b/d mice (27). Upon anti-Vγ2 TCR stimulation or culture with irradiated TL^b+ spleen cells as stimulators, Tg^b/d neonatal thymus graft–derived Vγ2⁺ IEL proliferated poorly and produced nearly undetectable levels of IL-2 (data not shown). Furthermore, unlike Vγ2⁺ IEL from Tg^d/d mice, which readily mediated the cytotoxic killing of Jurkat T cells upon stimulation with anti-Vγ2 mAb, both Tg^b/d neonatal thymus graft–derived Vγ2⁺ IEL and Vγ2⁺ IEL from Tg^b/d mice exhibited very weak cytotoxicity against human Jurkat T cells (data not shown).

Figure 3.

Efficient homing of Tg^b/d Vγ2⁺ neonatal thymocytes to the intestinal epithelium requires the extrathymic presence of the TL^b Ag. FCM analysis of IEL and spleen cells from (top row) athymic C57BL/6 (self-Ag TL^b+) mice and (bottom row) athymic BALB/c (TL^d+) mice that were grafted 6 weeks previously with the Tg^b/d (day 2) neonatal thymus. (middle row) FCM analysis of IEL and spleen cells from athymic C57BL/6 (self-Ag TL^b+) mice that were grafted with the Tg^b/d (6-week-old) adult thymus. Donor H-2⁺ cells represent cells derived from the neonatal thymus graft as determined by the appropriate anti–H-2K mAb (anti–H-2K^b or anti–H-2K^d). Data shown represent 1 of 3 independent experiments.

To determine whether the ability of Tg^b/d Vγ2⁺ neonatal thymocytes to escape negative selection and migrate to the intestinal epithelium is a consequence of absent or attenuated negative selection during the neonatal period, nude TL^b+ hosts were grafted with Tg^b/d adult thymus, in which the negative selection of Tg Vγ2⁺ is almost complete (Figures 1 and 2). Figure 3 (middle row) demonstrates that the adult Tg^b/d thymus generated graft-derived cells in both the IEL and spleen populations of nude TL^b+ hosts, but relatively few were Vγ2⁺ T cells. Although it is possible that the adult Tg thymus simply could not generate Vγ2⁺ IEL, this is unlikely because grafting of Tg^d/d adult thymus into TL^d+ BALB/c nude mice readily generated Vγ2⁺ IEL (data not shown). Overall, these results suggest that it is the absence or attenuation of negative selection during neonatal development that allows Tg^b/d neonatal thymus to generate Vγ2⁺ IEL in TL^b+ nude hosts.

Next, we examined whether the extrathymic presence of the TL^b Ag was required for the generation of Vγ2⁺ IEL in the nude host by grafting the Tg^b/d neonatal thymus into TL^d+ BALB/c nude mice, which do not bear the TL^b Ag. In this model, cells derived from the thymus graft (donor) can be distinguished from cells derived from the nude host by a mAb that specifically recognizes the donor-specific class I H-2K^b molecule (shown as donor H-2⁺ in Figure 3). Although Tg^b/d neonatal thymus readily generated T cells in the spleen and IEL populations of TL^d+ hosts, only a relatively small percentage of those T cells was Vγ2⁺ (Figure 3, bottom row). Although the total percentage of Tg^b/d neonatal thymus graft–derived Vγ2⁺ T cells found in the spleen and IEL of athymic TL^d+ hosts was very small, it constituted almost 10% of the total number of Tg^b/d neonatal thymus graft–derived T cells. This suggests that in the absence of the extrathymic TL^b Ag, autospecific Tg^b/d Vγ2⁺ neonatal thymocytes are capable of escaping negative selection and surviving in both the spleen and the intestinal epithelium at very low levels. However, as shown in Figure 3 (top row), when the Tg^b/d neonatal thymus was grafted into TL^b+ nude mice, the extrathymic presence of the TL^b Ag appeared to induce peripheral deletion of Tg^b/d Vγ2⁺ T cells in the spleen, and paradoxically, survival and increase in number of Tg^b/d Vγ2⁺ T cells in the intestinal epithelium. Overall, these results suggest that the intestinal epithelium of TL^b+ mice, but not TL^d+ mice, can serve as a sanctuary for Tg^b/d Vγ2⁺ neonatal thymocytes that have escaped negative selection and peripheral deletion.

Negative selection of Vγ2⁺ thymocytes is present but attenuated in the neonatal Tg^b/d thymus.

If neonatal Tg^b/d thymus is a major source of Tg^b/d Vγ2⁺ IEL, as suggested by our results, then an important question is whether negative selection of autospecific Tg^b/d Vγ2⁺ T cells is attenuated or simply absent in the neonatal thymus. As shown in Figure 4a, the total number of Vγ2⁺ thymocytes in the Tg^b/d neonatal thymus was reduced by 50% when compared with Tg^d/d neonatal thymus, suggesting that some form of negative selection is present in the neonatal Tg^b/d thymus. However, because there was a decrease of more than 95% in the total number of Vγ2⁺ thymocytes found in the adult Tg^b/d thymus compared with adult Tg^d/d thymus (data not shown), these results also suggest that negative selection of Vγ2⁺ thymocytes in the neonatal Tg^b/d thymus is present, but attenuated.

Figure 4.

(a) Negative selection of Tg^b/d Vγ2⁺ neonatal thymocytes is attenuated in day 2 neonatal thymus. The total number of Vγ2⁺ neonatal thymocytes is reduced by approximately 50% in Tg^b/d neonatal thymus when compared with Tg^d/d neonatal thymus. Data are shown as the mean of more than 20 neonatal thymus lobes examined separately for each group. (b) The TL^b self-Ag is functionally present in the neonatal Tg^b/d thymus. To test for the functional presence of the TL^b Ag, 10⁵ irradiated stimulator cells were cultured with adult Tg^d/d thymocytes that were sorted to remove TCR β cells to avoid a nonspecific allogeneic response. Data shown represent 1 of 2 independent experiments.

TL^b Ag is present in the Tg^b/d neonatal thymus.

We next explored the question of whether the attenuated level of negative selection in the Tg^b/d neonatal thymus may simply be the result of lower TL^b Ag expression in the neonatal thymus. To test this hypothesis, we looked for the functional presence of the TL^b Ag by testing the ability of irradiated Tg^b/d neonatal thymocytes to induce the proliferation of Tg^d/d Vγ2⁺ adult thymocytes, which should specifically recognize the TL^b Ag. To avoid an allogeneic response, we sorted and removed TCR β thymocytes before culturing with irradiated Tg^b/d neonatal thymocytes. Hence, any proliferative response should be the result of Tg^d/d Vγ2⁺ thymocytes recognizing the TL^b Ag. Figure 4b demonstrates that the TL^b Ag is functionally present in day 2 neonatal Tg^b/d thymus at levels equal to if not higher than those found in the adult Tg^b/d thymus. As expected, no functional TL^b Ag was detectable in the Tg^d/d adult thymus. Overall, these results suggest that the attenuation of negative selection of Tg^b/d Vγ2⁺ neonatal thymocytes is unlikely to be due to a lower level of TL^b Ag expression in the Tg^b/d neonatal thymus.

Phenotypic examination suggests that Tg^b/d Vγ2⁺ neonatal thymocytes have encountered the TL^b Ag at an immature stage in development.

If negative selection of Tg^b/d Vγ2⁺ neonatal thymocytes is partially functional and the TL^b Ag is present in the neonatal thymus, we rationalized that there should be some evidence that Vγ2⁺ Tg^b/d neonatal thymocytes have encountered the TL^b self-Ag. Comparison of Tg^b/d and Tg^d/d Vγ2⁺ neonatal thymocytes for several phenotypic markers of differentiation and activation revealed that Tg^d/d Vγ2⁺ neonatal thymocytes are comprised of a heterogeneous population expressing various levels of CD8α, CD45RB, CD44, and HSA. In contrast, Tg^b/d Vγ2⁺ neonatal thymocytes are a relatively homogenous population and are uniformly CD8^–, CD45RB^hi, CD44^hi, and HSA^d+ (Figure 5). Because Tg^b/d Vγ2⁺ neonatal thymocytes, but not Tg^d/d Vγ2⁺ neonatal thymocytes, are virtually all positive for CD44 and HSA, which have been reported to be markers of activation and thymocyte immaturity (36–39), these results suggest that Tg^b/d Vγ2⁺ neonatal thymocytes have been activated (presumably by the TL^b Ag), but are also blocked at an immature stage of development. Furthermore, closer phenotypic examination revealed that Tg^b/d Vγ2⁺ neonatal thymocytes express significantly lower levels of Thy 1 and Vγ2⁺ TCR than do naive Tg^d/d Vγ2⁺ neonatal thymocytes (Figure 6a). Therefore, we tested the hypothesis that lower levels of Thy 1 and TCR expression are markers of TL^b Ag recognition by examining naive Tg^d/d Vγ2⁺ neonatal thymocytes before and after activation in vitro with either anti-Vγ2 mAb or irradiated TL^b+ Tg^b/d spleen cells. Figure 6b demonstrates that, consistent with our hypothesis, activation of Tg^d/d Vγ2⁺ neonatal thymocytes resulted in lower levels of Thy 1 and Vγ2 TCR expression.

Figure 5.

FCM analysis of neonatal Tg^b/d (left) and Tg^d/d (right) thymocytes for markers of T-cell development and activation. Data shown represent 1 of 3 independent experiments.

Figure 6.

Phenotypic examination of Thy 1 and Vγ2⁺ TCR levels suggests that most Tg^b/d Vγ2⁺ neonatal thymocytes have recognized the TL^b Ag. (a) Tg^b/d Vγ2⁺ neonatal thymocytes express lower levels of Thy 1 and Vγ2⁺ TCR than do Tg^d/d Vγ2⁺ neonatal thymocytes. (b) Stimulation of Tg^d/d neonatal thymocytes with anti-Vγ2 TCR mAb resulted in decreased levels of Thy 1 and Vγ2⁺ TCR. Data shown represent 1 of 2 separate experiments.

Discussion

Using G8 γδ Tg mice, we have demonstrated that the negative selection of γδ T cells bearing an autospecific TCR is markedly attenuated in the neonatal thymus. A consequence of this attenuated negative selection is that potentially self-reactive thymocytes are allowed to escape negative selection, and migrate to the intestinal epithelium where they survive and increase in number. Although several studies have shown that negative selection is either absent or attenuated in the neonatal thymus, our study is the first to demonstrate that the fate of autospecific neonatal thymocytes can be other than death by negative selection. What makes our results intriguing is that we have also demonstrated that despite our autospecific neonatal γδ thymocytes appearing to be in the process of undergoing negative selection, the intestinal epithelium somehow serves as a sanctuary for these autospecific neonatal thymocytes. Lessons learned from studying this pathway are relevant to our understanding of how autoreactive T cells are allowed to develop and persist in autoimmune disease. It is worth noting that most studies that have addressed the fate of autospecific T cells have done so by transferring naive autospecific T cells into a host that expresses an Ag that is recognized by the transferred autospecific T cells. These studies, however, may not provide accurate models for studying the fate of autoreactive T cells in autoimmune disease, because they assume that autoreactive T cells are naive and have never encountered the self-Ag in development (i.e., negative selection). Our study addresses this issue and specifically examines the fate of autospecific T cells that have truly escaped functional negative selection.

From our model, 3 criteria appear to be required for autospecific thymocytes to escape negative selection and find sanctuary. The first criterion is that negative selection must be attenuated so that thymocytes expressing an autospecific TCR are allowed to escape negative selection. In our model, the neonatal period provides a brief window of time during which autospecific thymocytes can escape negative selection. It would be interesting to speculate whether stress or infection (which are both associated with autoimmune disease) can induce a similar phenomenon. The second criterion is that the self-Ag must be present extrathymically. Grafting of the Tg^b/d neonatal thymus generated a large number of Vγ2⁺ T cells in the intestinal epithelium of only Ag⁺ TL^b+ hosts, but not in Ag^– TL^d+ hosts (Figure 3). This criterion is puzzling, because our results suggest that Tg^b/d Vγ2⁺ neonatal thymocytes have already encountered the TL^b Ag and are rendered unresponsive to TCR stimulation (data not shown). The third criterion appears to be the environment. In our model, grafting of the Tg^b/d neonatal thymus into TL^b+ null hosts generated a large population of Vγ2⁺ T cells only in the intestinal epithelium and not elsewhere (Figure 3 and data not shown). This suggests that the intestinal epithelium provides a unique environment that allows the survival and expansion of autospecific neonatal thymocytes. The nature of this environment, however, is not presently known.

We also attempted to determine whether encountering the TL^b Ag intrathymically early in ontogeny gives Vγ2⁺ neonatal thymocytes a selective advantage in their ability to migrate to the intestinal epithelium of TL^b+ nude mice. Unfortunately, neither injection of Tg^d/d Vγ2⁺ thymocytes nor grafting of the Tg^d/d neonatal thymus into TL^b+ nude mice generated any graft-derived cells in the IEL, spleen, or lymph node populations of the nude host (data not shown). This is probably because host natural killer cells reject thymus graft–derived T cells in a fully allogeneic model. On the other hand, grafting of Tg^d/d neonatal thymus into a syngeneic TL^d+ nude host readily generated Tg Vγ2⁺ IEL (data not shown). This suggests that, at the very least, encountering the TL^b self-Ag in the neonatal thymus is not necessary for Tg Vγ2⁺ neonatal thymocytes to migrate to the intestinal epithelium.

An important question is whether our observations are also applicable to autospecific neonatal αβ thymocytes. Studies on αβ T cells, which recognize an endogenously expressed Mls Ag, and Tg αβ T cells, which recognize the male HY Ag, have shown that autospecific αβ T cells are found in relatively large numbers in both the neonatal thymus and IEL populations (8–11, 17, 20). These findings suggest that reminiscent of our observations regarding autospecific neonatal γδ thymocytes, the intestinal epithelium can also serve as a sanctuary for autospecific neonatal αβ thymocytes. However, whether these thymocytes can migrate to the intestinal epithelium and find sanctuary has not been directly demonstrated and is presently under investigation.

We have shown that Tg^b/d Vγ2⁺ neonatal thymocytes can give rise to Tg^b/d Vγ2⁺ IEL. Interestingly, the phenotypes of these 2 lineage-related populations are markedly different. For example, most Tg^b/d Vγ2⁺ neonatal thymocytes are HSA⁺, B220^–, CD8^– , Thy 1^lo, whereas most Tg^b/d Vγ2⁺ IEL are HSA^–, B220⁺, CD8⁺, Thy 1^– (Figure 5 and data not shown). This suggests that several additional developmental steps are required before Tg^b/d Vγ2⁺ neonatal thymocytes can develop into Tg^b/d Vγ2⁺ IEL. Whether these steps take place intrathymically or extrathymically (presumably at the intestinal epithelium) is unclear. The absence of a discrete population of Tg^b/d Vγ2⁺ neonatal thymocytes that phenotypically resemble Tg^b/d IEL suggests that despite their thymic origin, Tg^b/d Vγ2⁺ neonatal thymocytes differentiate further extrathymically before becoming Tg^b/d Vγ2⁺ IEL. However, we can not rule out the possibility that a very small portion of Tg^b/d Vγ2⁺ neonatal thymocytes differentiates intrathymically before leaving the thymus and migrating to the intestinal epithelium. Although this hypothesis is speculative, it may explain why the injection of Tg^b/d neonatal thymocytes was significantly less efficient at generating Tg^b/d Vγ2⁺ IEL in the TL^b+ nude host than was grafting of the neonatal Tg^b/d thymus (compare Figure 2b and Figure 3, top row).Our results with neonatal thymectomy and thymus grafting strongly suggest that most (if not all) Tg^b/d Vγ2⁺ IEL are derived from the thymus. Thymectomy on day 2 almost completely eliminated Tg^b/d Vγ2⁺ IEL (Figure 1c), and grafting of the neonatal Tg^b/d thymus into a TL^b+ host generated very large numbers of Tg^b/d Vγ2⁺ IEL (Figure 3, top row). At first sight, our results appear to directly conflict with the study by Barrett et al. (16), who used the adult thymectomy radiation bone marrow chimera (ATXBM) as a model for extrathymic T-cell development and concluded that Tg^b/d Vγ2⁺ IEL are derived from an extrathymic pathway. Although we are willing to concede that some Tg^b/d Vγ2⁺ IEL can develop through such a pathway, several lines of reasoning suggest that the results of the ATXBM model probably exaggerate the significance of the this pathway. First, the radiation bone marrow chimera model only demonstrates that some Tg^b/d Vγ2⁺ IEL can develop through an extrathymic pathway. Second, the ATXBM model is not natural and does not duplicate normal ontogeny as well as the thymus graft and neonatal thymectomy experiments do. Lastly, radiation probably alters the intestinal epithelium so that it induces the appearance of T cells, which normally would not be there. To prove this, we injected Tg^b/d spleen cells (which have virtually no Vγ2⁺ T cells) into a TL^b+ RAG null host and obtained virtually no Tg Vγ2⁺ IEL, as expected. However, when we irradiated the TL^b+ RAG null mice with 400 rads before injection of Tg^b/d spleen cells, there was an almost 30-fold increase in Tg Vγ2⁺ IEL in the recipient TL^b+ RAG null host (Lin et al., unpublished observations). Our conclusion is supported by the study of Rocha et al., who demonstrated that reconstitution of thymectomized, nonirradiated RAG null mice with bone marrow cells generated very few γδ IEL (40). Collectively, these results suggest that most of the γδ IEL seen in the ATXBM model probably would not appear under normal nonirradiated conditions, and that most γδ IEL (including Tg^b/d Vγ2⁺ IEL) are derived from the thymus.

If Tg^b/d Vγ2⁺ neonatal thymocytes can escape negative selection and efficiently migrate to and increase in number at the intestinal epithelium, as suggested by our data, it is unclear what purpose this would serve in developing a functional immune repertoire. Because our data suggest that most Tg^b/d Vγ2⁺ IEL are derived from the neonatal thymus, it is not surprising that the phenotype and the functional properties of Tg^b/d neonatal thymus–derived Vγ2^± IEL are essentially the same as those of Tg^b/d Vγ2⁺ IEL. The latter have been extensively studied by Barrett et al. (16, 27), who demonstrated that Tg^b/d Vγ2⁺ IEL have acquired tolerance. Although the development of tolerance may protect the intestinal epithelium from being harmed by potentially autoreactive T cells, it is not clear what functional purpose would be served by expanding a large population of both tolerant and potentially autoreactive T cells. One possibility is that this pathway may be an efficient way of generating IEL that produce TH2-type cytokines rather than the TH1-type cytokines, which may be more harmful to the intestinal epithelium in an immune reaction. Using 2C Tg mice that express an αβ TCR that recognizes an endogenous peptide Ag presented by the MHC class I H-2Ld, Guehler et al. (41) demonstrated that Tg IEL from Ag⁺ mice proliferated poorly and produced very little IL-2 or IFN-γ, yet also demonstrated by RT-PCR that Tg IEL from Ag⁺ mice were more likely than Tg IEL from their experimental Ag^– mice to produce TH2-type cytokine IL-4. Whether this phenomenon occurs for αβ IEL in other Tg mice or for our Tg Vγ2⁺ IEL has yet to be shown. Finally, our results may have some relevance to human γδ IEL. Groh et al. recently demonstrated that some human γδ IEL are capable of recognizing stress-induced self MHC class I–related molecules (42). Their results suggest that a similar pathway of development, which we have described here, may also occur for human γδ IEL.