Evidence for the involvement of lung-specific γδ T cell subsets in local responses to Streptococcus pneumoniae infection

Although γδ T cells are involved in the response to many pathogens, the dynamics and heterogeneity of the local γδ T cell response remains poorly defined. We recently identified γδ T cells as regulators of macrophages and dendritic cells during the resolution of Streptococcus pneumoniae-mediated lung inflammation. Here, using PCR, spectratype analysis and flow cytometry, we show that multiple γδ T cell subsets, including those bearing Vγ1, Vγ4 and Vγ6 TCR, increase in number in the lungs of infected mice, but not in associated lymphoid tissue. These γδ T cells displayed signs of activation, as defined by CD69 and CD25 expression. In vivo BrdU incorporation suggested that local expansion, rather than recruitment, was the principal mechanism underlying this increase in γδ T cells. This conclusion was supported by the finding that pulmonary γδ T cells, but not αβ T cells, isolated from mice that had resolved infection exhibited lung-homing capacity in both naive and infected recipients. Together, these data provide novel insights into the origins of the heterogeneous γδ T cell response that accompanies lung infection, and the first evidence that inflammation-associated γδ T cells may exhibit distinct tissue-homing potential.

Introduction cd T cells are a rare population of T lymphocytes in most tissues. The cd TCR + population can be subdivided based on the expression of specific Vc and Vd TCR chains [1], defined here using the nomenclature of Heilig and Tonegawa [2]. The resultant clonal and oligoclonal subsets may preferentially associate with specific tissues [3]. For example, expression of the Vc5 chain is predominantly restricted to dendritic epithelial cd T cells found in skin [4] and Vc6 + cells are the major reproductive tract cd T cell population [5]. In contrast, heterogeneous populations of cd T cells are found in other tissues, including the lung.
Substantial evidence exists therefore to place cd T cells as central immunoregulatory cells not only within the lung but in many tissues [3,17]. Despite the importance of immunoregulation in generating appropriate immune responses, cd T cells remain relatively poorly characterized during responses to infection. During resolving Streptococcus pneumoniae-induced lung inflammation [18,19], cd T cell numbers increase >30-fold and are responsible for regulating macrophage and dendritic cell numbers during the resolution phase of inflammation [20]. Using RT-PCR, spectratyping, flow cytometry and analysis of cell cycle progression, these cd T cells are shown here to represent an expansion of multiple lung-resident cd T cell subsets and to have an activated phenotype. The response is lung restricted, and cd T cells isolated from resolving lungs preferentially home back to the lungs of recipient mice in an inflammation-independent manner. Together, these data provide new insights into the lung cd T cell population and their tissue-specific nature.

Multiple Vc chain-expressing cd T cell populations in naive and inflamed lungs
Data characterizing cd T cells during pathogen-induced responses remain sparse. We have used a model of resolving S. pneumoniae infection [18,19] to examine the pulmonary cd T cell response following local pathogen-induced inflammation [20]. In this model, intranasal challenge with S. pneumoniae serotype 6B had little effect on the numbers of lung CD4 + or CD8 + T cell populations (Fig. 1A) over the subsequent 14 days. In contrast, the pulmonary cd T cell population was significantly increased, with a >30-fold increase in cd T cell number observed at the peak of the response (days 7-10; Fig. 1A). While the cd T cell response subsequently subsided, it remained significantly elevated at day 14 compared with naive controls (Fig. 1A).
Given this striking change in the lung cd T cell population, we next addressed whether one or more specific cd T cell subsets were responding to S. pneumoniae challenge. Qualitative PCR analysis of highly purified lung cd T cells from naive animals (n = 4 pools of six mice) showed a mixed population, with mRNA for Vc1, Vc2, Vc4, Vc5 and Vc6 detectable (Fig. 1B). Analysis of cd T cells purified at day 7 post S. pneumoniae challenge (n = 6 pools of three to six mice) also showed expression of mRNA for Vc1, Vc2, Vc4, Vc5and Vc6. The Vc5 chain commonly associates with Vd1 in DETC, although expression of mRNA for Vd1 was barely detectable in cd T cells purified from naive lung. In contrast, mRNA for Vd1 was consistently detectable in samples from day 7 post challenge ( Fig. 1B). Vc7 mRNA was not detected in any samples from either naive or infected mice.
In contrast with the lung, sorted cd T cells from the draining LN of naive (n = 3 pools of three to six mice) and S. pneumoniae-challenged mice (n = 3 pools of three to six mice) expressed only Vc1, Vc2 and, weakly, Vc4 mRNA (Fig. 1B).

Expansion of Vc1 + and Vc4 + cd T cells following pneumococcal challenge
To confirm the PCR analysis and to obtain quantitative data regarding cd T cell subsets, flow cytometric analysis of Vc1, Vc4, Vc5 and Vd6.3 expression was carried out on total lung cells. In naive mice, Vc1 + cells represented a mean 12.1 AE 2.0% (n = 9) of total cd T cells. At day 7 following S. pneumoniae challenge, this proportion was maintained, with 13.5 AE 1.1% (n = 12) of cd T cells expressing Vc1 ( Fig. 2A). Accounting for the increased total viable lung cell number associated with inflammation, a significant increase in the mean number of Vc1 + cells results following pneumococcal challenge. The absolute number of pulmonary Vc1 + cells rose from 3.6 AE 0.6 Â 10 3 in naive mice, to 1.25 AE 0.1 Â 10 5 at day 7 (p <0.001; Fig. 2B), a 35-fold increase.
Vc4 + cells were also well represented among cd T cells in the naive lung (mean 21.5 AE 1.5%) ( Fig. 2A). While pneumococcal challenge resulted in a significant decrease in the relative proportion of Vc4 + cells at day 7 (12.3 AE 1.2%; p = 0.001), this represented an almost 20-fold increase in their absolute number, from a mean of 6.5 AE 0.4 Â 10 3 cells/lung in naive mice to 1.14 AE 0.1 Â 10 5 at day 7 (p <0.001; Fig. 2B).
Expression of the Vd6.3 chain, which can associate with Vc1 [21], was consistently restricted to a small proportion of naive lung cd T cells (mean 1.9 AE 0.3%), which was reduced to 0.3 AE 0.2% at day 7. While this translated to a fourfold increase in the mean number of these cells in the lung, they remained relatively infrequent at day 7 (mean 2.6 AE 1.4 Â 10 3 cells/lung). No distinct Vc5 + population was present in either naive animals or following S. pneumoniae challenge, as determined by flow cytometry. The increase in number of both the Vc1 and Vc4 subsets, as well as of the remaining undefined cd T cells, followed similar kinetics (Fig. 2B). That is, in each case, the majority of the increase occurred between days 5 and 7 post challenge and peaked at days 7-10.
In comparative stainings of naive draining LN (Fig. 2C), >75% of cd T cells expressed either Vc1 or Vc4. These results indicate that a significant proportion of lung cd T cells (>65% in naive and >75% in infection) remain unidentifiable by the current panel of antibodies, which is not due to an inability of the antibodies to recognize their specific TCR in our hands. PCR analysis ( Fig. 1B) suggests that these cells may comprise Vc2 + and/or Vc6 + cells, for which antibodies are not available. Nevertheless, together these results clearly demonstrate that S. pneumoniae challenge results in increased numbers of multiple cd T cell subsets in the lung.
Non-productive Vc5 transcripts are present in lung cd T cells The expression of Vc5 mRNA in naive and inflamed lungs was unexpected, since Vc5 + T cells are reportedly restricted to skin [4], with minor populations found in murine mammary gland [22] and in spleens of Vc1 -/mice [23]. However, expression of cell surface Vc5 protein was not detected by flow cytometry (Fig. 2A). We therefore carried out spectratyping analysis of PCR products to determine whether the Vc5 transcripts were putatively productive.
The canonical, oligoclonal dendritic epithelial cd T cell population has Vc5 spectratype lengths of 142 and 145 bp [23]. However, Vc5 transcripts from both naive cd T cells and cd T cells from S. pneumoniae-challenged mice comprised spectratypes ranging from 126 to 142 bp. The in-frame, 142-bp spectratype was among the least represented of those observed, with no apparent bias towards other putative productive, inframe transcripts (139, 136, 133 and 130 bp) compared with non-productive (out-of-frame) spectratypes (Fig. 3A). In contrast, analysis of Vc4 PCR products showed dominant (>90%) expression of the major productive (183 and 186 bp) spectratypes (Fig. 3B). Together, flow cytometry and spectratype data strongly suggest that Vc5-expressing T cells do not occur in naive lung, and are not induced following S. pneumoniae challenge. cd T cells exhibit a highly activated phenotype following S. pneumoniae challenge To determine whether lung cd T cells are activated as a result of S. pneumoniae challenge, activation-associated surface marker expression and inflammatory cytokine production were examined by flow cytometry. In naive lungs, significantly more cd T cells were CD44 HI (mean 72.6 AE 5.0%) than were ab T cells from the same individuals (22.6 AE 1.6%, p = 0.001). Similarly, 43.3 AE 4.9% of naive lung cd T cells expressed the early activation marker CD69, compared with only 2.1 AE 0.3%% of naive ab T cells (p = 0.004; Fig. 4A).
At day 7 following S. pneumoniae challenge, 92 AE 3% of cd T cells were CD44 HI , significantly increased from naive levels (p <0.01.). While CD44 HI ab T cells were also increased compared with naive controls (50.4 AE 1.9%; p <0.01), they remained significantly less represented than among cd T cells at the same time point (p <0.001; Fig. 4A). A similar pattern was observed for CD69 expression, where the proportions of both cd (83 AE 5%) and ab T cells (36.0 AE 4.1%) expressing CD69 at day 7 post challenge were increased compared to their naive counterparts (p <0.01 in both cases). However, CD69 expression among cd T cells remained more prevalent than among the ab T cell population in the lung (p <0.001; Fig. 4A). At day 7 following challenge, 75 AE 3% of cd T cells were CD44 + CD69 + , in comparison to only 27.2 AE 2.4% of ab T cells (p <0.001; Fig. 4B).
The expression of CD25, a marker of T cell activation in vitro while associated with a regulatory T cell phenotype in vivo [24], further delineated the two T cell populations. Whereas the vast majority (>95%) of cd T cells from naive lungs were CD25 -, cd T cells examined at day 7 following pneumococcal challenge showed expression of CD25. In comparison, naive lung ab T cells were CD25 -, and remained so following S. pneumoniae challenge (Fig. 4C).
Previous studies have associated cd T cell function with the production of soluble mediators [25], including cytokines [12,[26][27][28]. However, ex vivo intracellular flow cytometry performed without exogenous stimuli did not detect expression of TNF, IFN-c, IL-12/23 p40, IL-10 or IL-4 within cd T cells from naive mice, or from mice at days 1, 3, 7, 10 and 14 post challenge. Quantitative real-time RT-PCR analysis of mRNA from purified, non-stimulated cd T cells further confirmed no consistent change in cytokine mRNA accumulation (data not shown).
To determine whether the commitment of lung cd T cells for cytokine production changed as a consequence of inflammation, lung cells were incubated in the presence of PMA and ionomycin prior to staining for cytokine expression. Given this stimulation, a high proportion of lung cd T cells from naive mice expressed IFN-c (35.4 AE 7.6%; Fig. 5), with few cells expressing detectable levels of IL-10 (1.3 AE 0.4%) or IL-4 (4.3 AE 1.8%). During the peak of the response, at day 10, the proportion of IFN-c-positive cd T cells was significantly reduced to 17.7 AE 3.4% (p = 0.04; n = 5; Fig. 5). Given the overall 30-fold increase in lung cd T cell number, these data indicate quantitative increases in cd T cells with IFN-c-producing capability at this stage of the response (mean 1.8 AE 0.3 Â 10 5 IFN-c + cells) compared with naive mice (mean 0.1 AE 0.02 Â 10 5 ). In contrast to cd T cells, the proportion of IFN-c-positive CD4 + T cells was increased at this time point (12.5 AE 4.2%) compared with naive mice (7.1 AE 1.8%; p = 0.05) Among cd T cells, the proportion of IL-10-(0.5 AE 0.2%; p = 0.06) and IL-4positive (1.7 AE 0.9%; p = 0.12) cd T cells also appeared to be somewhat, though not significantly, reduced. Together, these data demonstrate lung cd T cell activation as a result of S. pneumoniae challenge, with elevated expression of activation markers occurring in the absence of significant increases in the proportion of cells capable of expressing common effector cytokines.

Localized cell division of lung cd T cells
Despite clear increases of cd T cell numbers during inflammatory responses, the role of local expansion versus recruitment in this process has been implied [10,11,13] but not directly addressed. It was of significant interest to determine whether S. pneumoniae-induced cd T cells arise through recruitment to the lung from other sites or by expansion of resident lung cd T cells. Our flow cytometric data demonstrated no prominent increase or decrease in cd T cell numbers in the spleen, blood and draining LN of S. pneumoniae-challenged mice, compared with lung (data not shown), supporting a model in which S. pneumoniae-induced inflammation results in expansion of a previously resident, lungspecific cd T cell population. Therefore, lung ab and cd T cells were examined to determine whether these populations were undergoing expansion through division.
In naive mice, similar percentages of lung cd (12.8 AE 2.5%) and ab T cells (9.1 AE 3.0%) incorporated BrdU over a 7-day labeling period (Fig. 6). At day 7 following S. pneumoniae challenge, BrdU + cd T cells were significantly increased compared with naive mice (42.2 AE 8.1%; n = 6, p <0.01). Furthermore, lung cd T cell turnover at this time point was significantly greater (p <0.05) than that of lung ab T cells (31.2 AE 4.1%; n = 6) (Fig. 6). Examination of cd T cells from the spleen and draining LN of mice at day 7 post challenge did not reveal any significant increase in BrdU incorporation compared with naive controls (p >0.05; data not shown).

Pulmonary cd T cells have lung-homing capabilities
Tissue-specific association of cd T cells may be reflected in their trafficking abilities, with Vc5 + cells most clearly shown to have skin-homing capability [29,30]. To examine whether cd or ab T cells from S. pneumoniaechallenged mice were able to traffic specifically to the lung, day 7 cd and ab T cell populations from CD45.1 + donors were labeled with CFSE and transferred into naive CD45.2 + recipients. At 48 h post transfer, CD45.1 + CFSE + donor cd T cells preferentially located in the lung, where they comprised a significantly greater proportion of total CD3 + cells than in other organs examined (Fig. 7A). Accounting for differences in CD3 + cells in each organ, it is estimated that 40% of the recovered CD45.1 + CFSE + donor cd T cells were found in the lung, with approximately 50% in the spleen. In contrast, only 10% of the recovered ab T cells were found in the lung, with more than 85% associated with the spleen. This preferential lung-specific localization also occurred following S. pneumoniae challenge. Thus, transfer at day 3 post S. pneumoniae challenge resulted in a similar distribution of CD45.1 + CFSE + donor cd T cells at 1 or 4 days post transfer (day 4 or day 7 post challenge for recipients) (Fig. 7A). In contrast, ab T cells purified from lungs of the same donors did not show preferential lung homing. Donor ab T cells were detected at equivalent proportions in all organs of both naive and infected recipients that were examined (Fig. 7B). Again, accounting for differences in CD3 + cells in each organ at 4 days post transfer, over 50% of CD45.1 + CFSE + donor cd T cells were found in the lung, and only 40% in the spleen. For ab T cells, approximately 15% of the recovered cells were found in the lung, with 65% associated with the spleen. This suggests that tissue-associated inflammatory signals are not required for, but may enhance, homing of lung-derived cd T cells.

Discussion
Using an S. pneumoniae-induced model of pulmonary inflammation, this study has examined in detail the cd T cell population involved in the ensuing immune response. A substantially increased pulmonary cd T cell population arises as a result of pneumococcal challenge [20], and here we demonstrate the coincident involvement of multiple cd T cell subsets. While a significant proportion of naive lung cd T cells exhibited an activated phenotype, activation was clearly enhanced in infected mice, most notably in respect to CD25 expression. Finally, increased BrdU uptake and the preferential lung-homing capability of lung cd T cells following pneumococcal challenge strongly suggest that local division, rather than recruitment, is responsible for this infection-induced increase in cd T cell number.
Intranasal S. pneumoniae serotype 6B challenge induced a >30-fold increase in pulmonary cd T cells, without compensatory or complementary changes within draining LN, blood or spleen. This expanded population of cd T cells has cytotoxic activity against alveolar macrophages and lung dendritic cells, and acts to regulate these populations during resolution of inflammation [20]. However, in this and other infection models, immunoregulatory functions have been ascribed to bulk cd T cells without detailed analysis of the receptor repertoire of the responding populations [12,15,16,20,28,31]. Here, we have quantified Vc1 + and Vc4 + cells in naive lungs, confirming and extending recent immunohistochemical studies [8]. Flow cytometric analyses were unable to account for a significant proportion of lung cd T cells in both naive and inflamed lungs, in contrast to those in draining LN, which may include pulmonary Vc6 + cells [8]. While Vc6 + cd T cells have recently been identified by application of the anti-Vc5 mAb 17D1 following staining with an antibody against the cd T cell receptor [6,8], this technique has not been successful in identifying Vc6 + cells in our laboratory. However, PCR data support the indirect observation of pulmonary Vc6 + cells [8]. While this suggests that both Vc6 + and Vc2 + cells are likely to comprise any remainder of the cd T cell population, a majority of Vc2 transcripts may be non-productive, as indicated by spectratype data [23]. Further analysis of the response did not reveal any apparent differences in the kinetics of each responding subset. This suggests that, in contrast with experimental infections such as listeriosis or schistosomiasis [32][33][34], there is no initial cd T cell response associated with anti-bacterial function. In contrast, the entire response appears to correlate with the final stages of bacterial clearance and the onset of the resolution phase.
PCR analysis of naive and challenged mice also revealed lung-restricted expression of Vc5 mRNA. While Vc5 + cells are not completely restricted to the skin dendritic epithelial T cell population [17,22,23], pulmonary Vc5 mRNA was unexpected. However, neither spectratype analysis nor flow cytometry, in agreement with recent data [8], indicated the presence of any currently defined Vc5 + population. Non-productive Vc5 mRNA may be erroneously expressed by, for example, Vc6 + cells, as suggested by previous studies [35][36][37]. Therefore, lung cd T cells expressing functional Vc5 TCR appear to have been excluded.
Vc1 + and Vc4 + cells were significantly increased following S. pneumoniae challenge. Pulmonary Vc1 + and Vc4 + cells have previously been shown to respond to both mycobacteria [13] and influenza [11]. While cd T cell cytokine production is observed in both models, only mycobacterial challenge induces cd T cell cytotoxicity [12,13,28]. In contrast, pneumococcal challenge induces resolution-associated cytotoxic activity [20] without coincident ex vivo cytokine expression. Stimulation of cd T cells with PMA and ionomycin revealed no overall change in the capacity of these cells to produce specific cytokines following challenge. While these data suggest increased numbers of cd T cells with cytokineproducing potential during the inflammatory response, any role for cd T cell-derived cytokines in the current model remains to be defined.
It remains unclear whether Vc1 + and Vc4 + populations recognizing similar targets are expanded in these disparate contexts, suggestive of functional heterogeneity dependent on an inflammatory environment rather than the ligand encountered. Alternatively, each inflammatory situation may activate distinct populations of Vc1 and Vc4 cells, each with specific functions. Despite this first quantification of pulmonary cd T cell subsets during inflammation, a majority of responding cd T cells remain undefined by direct observation. Nevertheless, by implication from PCR analyses together with flow cytometric data, S. pneumoniae-induced inflammation clearly induces substantial quantitative increases in multiple cd T cell subsets.
The majority of cd T cells from naive lungs constitutively expressed the activation-associated markers CD44 and CD69. This apparently high level of constitutively activated pulmonary cd T cells fits within a model of cd T cell population maturation in naive animals [38]. Phenotypically activated, 'naive' cd T cells also occur in murine vaginal [39,40] and intestinal epithelium [41], but not the spleen of the same uninfected mice, or in recent thymic emigrants [38]. These data suggest that a mucosal location may closely associate with an activated status in apparently 'naive' hosts, although the relationship between activation and homeostatic function remains to be investigated. S. pneumoniae challenge further activated the multiple responding cd T cell subsets in a surprisingly uniform fashion. The activation response is not restricted to, for example, the Vc1 or Vc4 subset, as for Vc6 in both Listeria [6] and pulmonary fibrosis [10] models. In other systems, cd T cells may up-regulate surface markers following recognition of Toll-like receptor ligands [42], or as a result of TNF activity [43], although the mechanisms driving activation of cd T cells both in naive lungs and following in vivo pathogen challenge remain unknown.
Given the tissue-specific nature of certain cd T cell subsets, it may be assumed that increases in lung cd T cells arise from local expansion. While data concerning canonical Vc subsets, including Vc6 in the lung [10], strongly support this conclusion, other cd T cell subsets have widespread distributions. Thus, while increased Vc1 and Vc4 numbers may represent local expansion, these cells may feasibly migrate to the lung from draining LN, spleen and other lymphoid tissues. The current data are the first to directly examine this alternative, and indicate that local expansion is the primary source of increased numbers during a polyclonal cd T cell response. Following S. pneumoniae challenge, cell division among cd T cells is significantly increased in the lung, but not in draining LN or spleen, and cycling of lung cd T cells has been observed in response to influenza infection [44]. The inflammatory environment of the challenged lung may drive both activation and expansion of resident cd T cells. The lack of significant cd T cell division, or their increase or loss in other tissues and blood, supports this model. However, any potential contribution of thymic emigrant cells, while of a non-activated phenotype in naive animals [38], remains to be determined under inflammatory conditions. While the proportion of cycling (BrdU + ) lung ab T cells was also increased following challenge, this is not accompanied by substantial increases in their number. In contrast to the apparent local expansion of cd T cells, it is more likely that ab T cells expand in draining LN and home back to lung tissues. The absence of substantial quantitative changes suggests that the current model of rapidly resolving inflammation does not facilitate the prolonged retention or recruitment of ab T cells within the lung.
cd T cells from lungs of mice challenged 7 days previously with S. pneumoniae exhibit a strong preference in homing to the lung, regardless of the inflammation state of the tissue. This inflammation independence contrasts with the homing phenotype suggested from human in vitro studies [45], and may reflect the tissue, rather than blood, origin of the cd T cells. That lung-derived ab T cells from the same individuals did not exhibit this preference is further evidence of the tissue-specific nature of both the lung cd T cell response and of cd T cells in general. Two important considerations remain under examination with regard to homing capabilities. First, does each lung cd T cell subset have equal homing capacity? Further studies will examine whether lung homing is most associated with the canonical lung Vc6 subset. Second, is cd T cell homing to the lung restricted to cells that originate in the lung, or is it an inflammation-inducible property of cd T cells in other tissues? To address this latter point, we attempted to purify activated cd T cells from the spleen of Listeria-infected mice for transfer studies. However, it was not possible to purify sufficient splenic cd T cells due to their strong adherence to activated splenic macrophages in this model ( [46] and our unpublished data).
In vitro studies of human cd T cells suggest that these cells have LN-homing capabilities once activated [47], while others may acquire a tissue-homing phenotype [45]. It is possible that lung cd T cells may be 'imprinted' for lung homing either during development or activation, as occurs for ab T cells in the intestinal mucosa [48]. The mechanisms governing lung-specific cd T cell homing are the subject of ongoing study, although expression of the commonly mucosal lymphocyteassociated integrin CD103 by pulmonary cd T cells has not been observed (our unpublished observations).
Together these studies provide further evidence for the tissue-and context-specific nature of cd T cell responses. However, they also serve to highlight the paucity of knowledge concerning cd T cell behavior and its regulation. By further examination of the activation processes and tissue-specific regulatory mechanisms to which cd T cells are subject, a much better understanding of this important, immunoregulatory T cell population will emerge.

Materials and methods
Mice and S. pneumoniae infection C57BL/6 (B6.CD45.2) and B6.CD45.1 mice were bred and housed under barrier conditions at LSHTM and the University of York, and supplied with food and water ad libitum. During in vivo labeling experiments, drinking water was supplemented with BrdU (Sigma) at 0.8 mg/mL, freshly prepared daily. Mice were infected intranasally with approximately 10 8 CFU of S. pneumoniae serotype 6B as described [18,19]. Animal experimentation was performed with LSHTM and the University of York Animal Procedures Ethics Committee and UK Home Office approval.

Tissue preparation
Whole organs were collagenase-digested to single-cell suspensions as described [18]. Viable cell counts were determined by Trypan blue exclusion.
For flow cytometric sorting, whole lung preparations were enriched for T cells using biotinylated anti-CD3 antibody, antibiotin beads and MACS sorting (Miltenyi Biotech, Germany). Subsequent purification of cd and ab T cells was carried out on a MoFlo cell sorter (DakoCytomation). Purity of sorted populations was confirmed by flow cytometry as >95% in each case.

Adoptive transfer
Sorted cd or ab T cells were labeled by incubation in 5 lM CFSE for 10 min at room temperature, washed extensively and resuspended in PBS. Mice received 10 6 donor cells intravenously.

PCR and spectratype analysis
RT-PCR analysis of Vc chain expression, and subsequent spectratyping of PCR products, was carried out on mRNA extracted from whole organ preparations, or from sorted cell populations derived from pooled organs of three to six individuals, as described [23].

Statistical analysis
Where indicated, data were compared using Student's twotailed t-test.