Thymocyte-Dendritic Cell Interactions near Sources of CCR7 Ligands in the Thymic Cortex¹

Developing T cells travel between distinct anatomical compartments as they mature and undergo positive and negative selection in the thymus (reviewed in Ref. 1). In the outer portion of the thymus, termed the cortex, CD4⁺CD8⁺ thymocytes undergo positive selection based on the ability of their newly formed TCRs to recognize self peptide-MHC ligands expressed by cortical thymic epithelial cells. After positive selection, thymocytes migrate to the central medullary region, where they undergo screening for negative selection based on the recognition of self peptide-MHC expressed by dendritic cells (DCs)³ and medullary thymic epithelial cells (2, 3, 4). The cortical/medullary compartmentalization of the thymus is reflected in distinct cell types found in each region, as well as the distinct chemokine expression patterns. In particular, the ligands for the chemokine receptor CCR7 are preferentially expressed in the medulla and are thought to direct the migration of thymocytes from the cortex to the medulla following positive selection (5, 6).

To what extent the cortex itself is divided into distinct functional compartments is not yet clear. At one end of the spectrum, it is possible that the cortex is functionally homogeneous. According to this view, cortical thymocytes would invariably be in contact with thymic epithelial cells capable of mediating positive selection, and positive selection could take place with equal probability throughout the cortex. On the other hand, there are indications that cortical thymic epithelial cells are heterogeneous (reviewed in Ref. 7), raising the possibility of functionally distinct regions of the cortex specialized for mediating selection events. The anatomical segregation of thymocytes in the cortex based on their TCR specificity provides an indication for such heterogeneity (8). However, distinct functional regions of the thymic cortex have not yet been identified.

Two-photon laser scanning microscopy (TPLSM) provides an opportunity to examine the cellular migration and interaction events in tissues that underlie selection events within the thymus (reviewed in Ref. 9). Initial studies have revealed that thymocytes undergo a calcium-dependent stopping in response to positive selection signals in the thymus (10) and that interaction of thymocytes with MHC-bearing stromal cells can lead to prolonged but dynamic cell-cell contacts (11). Additionally, analysis of cell migration in the cortex of intact thymic lobes provided evidence that thymocytes move via a random walk in the cortex before positive selection and undergo rapid, directed migration to the medulla as a result of positive selection (12). These studies provide an early glimpse into the regulation of thymocyte migration by positive selection. However, the interactions of thymocytes with other cell types in the intact thymus and how these interactions change during the process of positive selection have not yet been addressed.

Herein we examine the impact of the chemokine receptor CCR7 and TCR repertoire selection on thymocyte-DC interactions in the cortex of intact thymic lobes. We show that thymic DCs form intimate associations with cortical capillaries near sources of CCR7 ligands in the cortex. We find that thymocytes extensively interact with DCs in the cortex and that thymocytes expressing TCRs that mediate positive selection via class I MHC have a greater frequency and increased duration of contact with DCs. We also show that expression of CCR7 on thymocytes correlates with their tendency to associate with cortical DCs. Taken together, our results identify a vasculature-associated microenvironment within the thymic cortex that brings together DCs and the thymocytes that are receiving positive selection signals. These structures may serve to promote cell-cell interactions involved in TCR repertoire selection in the thymic cortex.

All mice were housed and bred in pathogen-free conditions at the American Association of Laboratory Animal Care-approved animal facility at the Life Sciences Addition of the University of California, Berkeley. All animal experiments were approved by the Animal Care and Use Committee of the University of California, Berkeley. C57BL/6 (B6) mice (The Jackson Laboratory) were crossed to CD11cYFP (yellow fluorescent protein) transgenic mice on a B6 background (13) for use as neonatal hosts to generate partial hematopoietic chimeras as described previously (12, 14). Bone marrow donors were either UBI-GFP transgenic mice (15), or UBI-GFP mice crossed to CCR7 transgenic mice (6), or UBI-GFP mice crossed several times to P14 TCR transgenic Rag2^−/− mice (16) (purchased from Taconic Farms), or F5 TCR rag2^−/− (17) (purchased from Taconic Farms) to generate P14 TCR or F5 TCR and UBI-GFP transgenic Rag2^−/− donors. P14 TCR, UBI-GFP transgenic mice were crossed to CCR7^−/− mice (18) to generate P14 TCR, UBI-GFP transgenic CCR7^−/− donors. Co-chimeras were generated using actin-cyan fluorescent protein (CFP) transgenic B6 donors (19). High-density chimeras were generated using adult CD11cYFP hosts that had been treated with 5-fluorouracil to eliminate cycling cells, and they were i.v. injected with bone marrow from P14 TCR and UBI-GFP transgenic Rag2^−/− donors and actin-CFP donors. CCR7^−/− and plt/plt mice (20) were provided by Jason Cyster (University of California, San Francisco).

At 4 to 6 wk of age, chimeric mice were sacrificed and individual thymic lobes were imaged by TPLSM while being perfused with warmed, oxygenated DMEM as described previously (12, 14). For two-color microscopy, imaging volumes (164 × 164 × 40 μm) corresponded to regions of the thymic cortex extending from 80 to 200 μm below the surface of the capsule and were scanned every 37 s for 20–40 min using a Zeiss 510 META/NLO two-photon microscope with a Spectra-Physics MaiTai laser. The distinct DC densities of the cortex and medulla were used to confirm that the imaging volumes corresponded to cortex. Two-photon excitation was achieved using a Spectra-Physics MaiTai laser tuned to 920 nm, and GFP and YFP emission light was separated using a 515 nm dichroic mirror and collected using two non-descanned detectors. For three-color microscopy, imaging volumes of 172 × 143 × 80 μm were generated using a custom-built microscope with a Spectra-Physics MaiTai laser tuned to 900 nm for excitation of CFP, GFP, and YFP and separation of emission spectra using 495- and 515-nm dichroics. Images of high-density chimeras were taken every 4 μm to make 80-μm z-stacks every 15 s for 20–30 min.

The x,y,z coordinates of individual thymocytes over time were obtained using Imaris Bitplane software, and motility parameters were calculated using either C or MATLAB (MathWorks; code available upon request). GraphPad Prism was used for graphing and statistical analysis. Speed was defined as path length divided by time (μm/min). For manual scoring of thymocyte-DC interactions, individual thymocyte tracks were inspected in three-dimensional space at each time point by an unbiased observer and scored for time points in which visible contacts with DCs occurred. Automated quantitation of thymocyte-DC interactions are described in detail elsewhere (Y. Chen, E. Ladi, P. Herzmark, E. Robey, and B. Roysam, manuscript submitted). Briefly, we first collected the k-nearest DC voxels (k = 5% total number of DC voxels) to each thymocyte instead of the single-nearest DC voxel and computed the average of Euclidean distances from the thymocyte center to each neighboring DC voxel. Then, we normalized the averaged center-to-edge distance by the thymocyte radii. Ultimately, the frequency of thymocyte-DC contacts is approximated by the number of time points at which the normalized distance drops below a user-defined threshold, against the total number of time points examined.

To determine the distribution of wild-type and P14 thymocytes relative to cortical DCs in high-density chimeras, DCs were delineated using a mean-shift clustering algorithm (Y. Chen, E. Ladi, P. Herzmark, E. Robey, and B. Roysam, manuscript submitted). For each x,y image, an area extending 5 μm outside the edge of the DC was defined as a DC interaction region, and the number of pixels corresponding to wild-type thymocytes (CFP) or P14 thymocytes (GFP) inside and outside of this DC interaction region relative to the total number of pixels was determined.

Isolated thymic lobes were fixed in 4% formalin and 10% sucrose for 1 h, sequentially submerged in 10, 20, and 30% sucrose for 10–18 h each, and frozen over dry ice in OCT. Serial sections (10–20 μm thick) were generated in-house as well as by BioPathology Sciences Medical. Sections were fixed with cold acetone for 10 min, blocked with 5% BSA in PBS for 30 min to1 h, stained with biotin-conjugated anti-Ly51 at 1/20 dilution (BD Biosciences), anti-CCL21 at 1/20 dilution of 50 μg/ml stock (R&D Systems), or anti-F4–80 at 1/20 dilution of 50 μg/ml stock (eBioscience) overnight at 4°C, and stained with streptavidin-Texas Red or Alexa 633 conjugate (Invitrogen) for 30 min at room temperature. To label vasculature, mice were injected via tail vein with 100–200 μg of tomato lectin-FITC or lectin-Texas Red (Vector Laboratories) just before sacrifice, as previously described (21). TUNEL labeling of thymic sections was performed according to manufacturer’s instructions using the in situ Cell Death Detection kit (Roche). Stained thymic sections were visualized on Zeiss 510 Axioplan META/NLO upright confocal microscope with a 40× oil objective (Plan-Neofluar 40×/1.3 oil WD = 0.17 mm) using 488- and 543-nm laser lines. Images were analyzed and assembled using Adobe Photoshop.

As an initial step to investigate the function of DCs in the thymus, we examined the thymic expression pattern of a reporter in which YFP is under the control of the DC-specific promoter CD11c (13). Analysis of fixed tissue sections of adult thymus from CD11cYFP mice showed that the reporter labels a population of cells with DC-like morphology (Fig. 1). Staining of thymic tissue sections for CD11c protein and flow cytometric analysis of collagenase-digested thymus revealed a good correspondence between the CD11c reporter and CD11c protein expression (data not shown). Although most thymic DCs were found in the medulla, there was also an extensive network of reporter-positive cells throughout the thymic cortex (Fig. 1 A).

Previous reports have shown that apoptotic cells (TUNEL⁺) in the thymic cortex are found predominantly within F4–80-positive macrophages (22). To investigate the relationship between cortical DCs and these phagocytic cells, we costained thymus sections of reporter mice with F4–80 or TUNEL (Fig. 1, B and C). Although we occasionally observe TUNEL⁺ cells in the cortex that appear to be enclosed by a reporter-positive cell (12 of 83, or 14%), the majority (56 of 83, or 67%) of apoptotic cells associate with the dendrites of CD11cYFP cells while some (15 of 83, or 18%) do not appear to contact reporter-expressing cells (Fig. 1,B). Additionally, while most of the F4–80 signal in the cortex does not colocalize with the CD11cYFP reporter (Fig. 1,C, upper panels), some reporter-positive cells express low, but detectable, levels of F4–80 (Fig. 1 C, lower panels). Taken together, these results suggest that the CD11cYFP reporter identifies a distinct population from the cortical macrophages described previously, which nevertheless do share some characteristics with macrophages. Based on their DC-like morphology and expression of the DC-specific marker CD11c, we refer to these cells as DCs.

Similar to the distribution of DCs reported herein, the chemokine CCL21, a ligand for CCR7, has also been reported to be concentrated in the medulla and present at lower levels throughout the thymic cortex (23, 24). To explore the relative locations of DCs and CCL21 in the thymus, we stained fixed sections of adult thymic lobes from CD11cYFP reporter mice with Abs specific for the CCR7 ligand, CCL21 (Fig. 2,A). We found that DCs and CCL21 have an overall similar pattern of distribution throughout the thymus, with the DCs often found directly adjacent to the CCL21 signal and not colocalizing with it. This pattern is observed both in the medulla, where CCL21 expression is highest (Fig. 2,A, lower left panel), and throughout the cortex, where a significant signal is also observed (Fig. 2 A, lower right panel).

It has been previously reported that CCL21 expression is associated with cortical capillaries (24). We confirmed this by examining fixed thymic tissue sections from mice injected with fluorescent lectin before sacrifice and stained with Abs for CCL21 (Fig. 2^,B). The specificity of CCL21 staining in this study was confirmed by either omitting primary CCL21 Ab (Fig. 2^,A, upper right panel) or by parallel staining of thymic sections of mice deficient for CCR7 ligands (plt/plt mice) (Fig. 2^,B, lower panels). We also directly examined the association of DCs with capillaries in the thymic cortex by injecting CD11cYFP reporter mice with fluorescent lectin and imaging intact thymic lobes by TPLSM (Fig. 2 C and supplemental video 1).⁴ We found that many cortical DCs formed intimate associations with capillaries (77 of 127, 60%), with dendrites aligned and in close contact with capillaries.

The distribution of DCs throughout the cortex suggests that CD4⁺CD8⁺ thymocytes have extensive opportunity to interact with DCs. To examine these interactions in real time, we generated partial hematopoietic chimeric mice in which ∼1% of thymocytes were derived from donor mice expressing a GFP transgene under the control of the ubiquitin promoter (UBI-GFP transgenic mice) (12, 15), and hosts expressed the CD11cYFP reporter (Fig. 3,A). We then used TPLSM to image the interactions between donor thymocytes and thymic DCs within the cortex of intact thymic lobes from adult chimeric mice. Consistent with thymus section data, we consistently observed between 10 and 50 DCs within imaging volumes of 164 × 164 × 40 μm within the thymic cortex (Fig. 3, supplemental videos 2–5, and data not shown). The vast majority of DCs imaged in this study remained immobile during the imaging period and exhibited small rapid movements of dendrites. However, we did observe a small number of reporter-positive cells (9 of 500, or ∼2%) that were actively migrating during the imaging period (Fig. 3 B and supplemental video 2).

To examine the impact of TCR repertoire selection on thymocyte-DC interactions in the cortex, we used the P14 TCR transgenic model (16). The P14 TCR is restricted to the MHC-I molecule D^b and leads to positive selection of CD8 T cells on the C57BL/6 (B6) background. Donor bone marrow from P14 TCR, GFP transgenic Rag2^−/− mice was injected into CD11cYFP transgenic neonatal hosts (P14 → CD11cYFP). Flow cytometric analysis of these chimeras confirmed that the donor-derived thymocytes had a similar CD4 and CD8 expression pattern as intact P14 TCR transgenic mice, with most GFP⁺ cells expressing both CD4 and CD8 (supplemental Fig. 1A). For comparison, we also generated chimeras using bone marrow from GFP transgenic mice that were otherwise wild type (wt → CD11cYFP). These partial hematopoietic chimeras provided model systems for observing thymocyte-DC interactions under conditions in which thymocytes express diverse polyclonal TCRs or in which positive selection predominates.

Examination of time-lapse images of intact thymic lobes from chimeric mice revealed an overall similar pattern of thymocyte-DC contacts regardless of whether thymocytes expressed polyclonal TCR or the positively selecting P14 TCR (Fig. 4 A and supplemental videos 4 and 5). The majority of cortical thymocytes engaged in frequent encounters with DCs, most of which consisted of brief contacts (<5 min). Many of these brief encounters occurred as thymocytes approached a DC and then rapidly changed direction upon encounter (supplemental video 6). We observed many examples of individual thymocytes making successive contacts with multiple DCs (supplemental videos 6 and 7). Contacts took place with the cell body of a DC (supplemental video 8) or with dendrite tips (supplemental video 9). Although most encounters were brief, we occasionally observed more long-lasting intimate contacts between thymocytes and DCs, including some that were maintained throughout the period of imaging (20–30 min) (supplemental video 8). Additionally, we found labeled thymocytes that appeared to be enclosed by a DC for several minutes (supplemental video 10), consistent with section staining showing occasional TUNEL⁺ cells inside cortical DCs.

From visual inspection of the data, we noted that thymocytes expressing the positive selecting P14 TCR (P14→CD11cYFP) appear to associate more closely with DCs compared with thymocytes expressing a diverse TCR repertoire (wt → CD11cYFP). To quantify the extent of thymocyte-DC association, we manually scored for cell contacts. For this analysis, we chose a set of runs in which the density of DCs was relatively low (10–30 DCs/field) to minimize the probability of random thymocyte-DC encounters. Individual thymocyte tracks were inspected in three-dimensional space at each time point and the number and duration of individual thymocyte-DC contacts were recorded (Table I, Figs. 4,B and 5). P14 TCR-expressing thymocytes were found in contact with DCs approximately three times as often as wild-type thymocytes. This correlated with a slight but significant (p < 0.05) increase in the average duration of the DC contacts (3.0 vs 2.2 min) and a >2-fold increase in the frequency of individual DC contacts (8.4 vs 3.3 contacts/h). While P14 cortical thymocytes contain a higher proportion of rapidly migrating cells (Fig. 5,A and Ref. 12), the increased DC association is observed even when thymocytes of intermediate speed are compared (Fig. 4,B, shaded area and Fig. 5,B, middle panel). To control for the effects of local tissue environment on thymocyte-DC interactions, we also examined DC interactions of P14 and wild-type thymocytes within the same imaging volumes using co-chimeras in which CD11cYFP transgenic hosts were injected with a mixture of bone marrow from P14 TCR (GFP) and wild-type (CFP) donors. Flow cytometric analysis of thymocytes of these chimeric mice showed that most CFP⁺ and GFP⁺ cells are CD4⁺CD8⁺ thymocytes (supplemental Fig. 1B). These analyses confirmed that P14 thymocytes make more frequent and lengthy contacts with DCs compared with wild-type thymocytes (Fig. 4, C and D, Table II, and supplemental video 11). Based on the frequency of thymocyte contacts reported herein, and previous estimates of 1–2 days for a thymocyte to remain as a CD4⁺CD8⁺ thymocyte (25, 26), we estimate that a typical thymocyte bearing a positive-selecting, class I-specific TCR contacts 100–400 DCs during its stay in the cortex.

To confirm these results, we also employed a more objective computational image analysis approach to quantifying cell-cell interactions. This approach, which is described in detail in a separate publication (Y. Chen, E. Ladi, P. Herzmark, E. Robey, and B. Roysam, manuscript submitted), delineates (segments) all the cells using a mean-shift clustering algorithm and computes the distance between the center of each thymocyte and the edge of the nearest DC. Since thymocytes with different diameters can have comparable distances, we normalized the distances by radii of the thymocytes. We compared DC associations between wild-type thymocytes and thymocytes expressing two different class I-restricted TCRs, P14 and F5 TCR (17), and defined a thymocyte-DC contact as a center-to-edge distance of less than the radius of the thymocyte. Using this approach, we estimate that wild-type thymocytes spend 2% of their time in contact with a DC, compared with 7.5% and 9% for P14 TCR and F5 TCR transgenic thymocytes, respectively. These analyses confirm the trend that greater DC associations correlate with positive selection, and extend this observation to a second class I-restricted transgenic TCR.

As an additional objective method for comparing DC interactions between wild-type and P14 thymocytes, we also defined a region of fixed size surrounding each DC and determined the ratio of thymocyte signal within the DC-proximal region to the thymocyte signal in regions that were not adjacent to a DC (Fig. 6, A and B). For these analyses, we used chimeras containing a high density of both labeled P14 (GFP) and wild-type (CFP) thymocytes, and we did not attempt to delineate or track individual thymocytes. GFP signal from P14-expressing thymocytes was found predominantly within the DC region, while a CFP signal corresponding to thymocytes expressing a diverse TCR repertoire was distributed equally inside and outside the DC region (Fig. 6 C). Taken together, our results show that thymocytes bearing positively selecting TCR associate more closely with DCs compared with thymocytes bearing polyclonal TCRs.

We have previously reported that positive selection correlates with a population of thymocytes that migrate rapidly and directionally toward the medulla (12). Interestingly, while the population that showed clear directional migration corresponded to the fastest cells (>13 μm/min), we find that thymocytes with intermediate speeds (6–13 μm/min), which do not show directional migration (12), interact extensively with DCs (Fig. 4,B, shaded area and Fig. 5 B, middle panel). Additionally, visual inspection of the tracks of thymocytes that move rapidly and directionally toward the medulla shows that they do not generally form close associations with DCs (supplemental video 12). Collectively, these data suggest that positive selection leads to the appearance of a population of cortical thymocytes with intermediate speed that associate closely with cortical DCs, and a distinct population that migrates rapidly toward the medulla.

To further explore the relationship between positive selection and thymocyte-DC associations, we also examined thymocytes from TCRα^−/− mice, which do not receive positive selection signals and are arrested at the CD4⁺CD8⁺ stage of development (27). TCRα-deficient thymocytes lack the population of intermediate motility cells with extensive DC interactions that are prominent among P14 TCR thymocytes (Fig. 4,B, shaded area), consistent with the view that these cells are generated by positive selection signals. However, the overall interaction of TCRα^−/− thymocytes with DCs was intermediate between P14 and wild-type thymocytes (Table I). We also noted that TCRα-deficient cortical thymocytes migrate more slowly and arrest more often compared with wild-type thymocytes (Fig. 5 A), consistent with our previous results with thymocytes expressing a transgenic TCR on a nonselecting background (12). These data suggest that cortical thymocytes, which either lack a TCR or express a TCR but lack relevant MHC ligands, differ in their behavior from cortical thymocytes that express polyclonal TCRs. This notion is also consistent with the observation that the bulk of CD4⁺CD8⁺ thymocytes from TCRα-deficient mice have slightly reduced levels of CD5 (a marker known to correlate with TCR signaling) when compared with CD4⁺CD8⁺ thymocytes from wild-type mice (Ref. 28 and supplemental Fig. 2A). Taken together, these results suggest that, in addition to lacking thymocytes with intermediate motility that interact extensively with DCs, TCRα-deficient thymocytes also contain a new population of “dead-end” cells that migrate more slowly and associate more closely with cortical DCs, when compared with thymocytes expressing polyclonal TCRs.

Positive selection induces a large set of changes in CD4⁺CD8⁺ thymocytes, including up-regulation of the chemokine receptor CCR7 (29, 30). For example, >20% of CD4⁺CD8⁺ thymocytes from P14 TCR transgenic mice are CCR7⁺ as compared with <5% of CD4⁺CD8⁺ thymocytes from nontransgenic mice (Ref. 31 and supplemental Fig. 2). Given the close association of DCs with cortical sources of CCR7 ligand (Fig. 2), we hypothesized that CCR7 expression on thymocytes that have begun to receive positive selection signals might promote associations with DCs.

To investigate this possibility, we asked whether enforced CCR7 expression could drive thymocyte-DC interactions in thymocytes bearing polyclonal TCR. For this, we used a transgenic model in which CCR7 is constitutively expressed on CD4⁺CD8⁺ thymocytes (6). We generated chimeric mice using bone marrow donor cells from GFP, CCR7 transgenic donors (CCR7tg → CD11cYFP), imaged intact thymic lobes from adult chimeric mice, and quantitated thymocyte-DC interactions (Table I and Fig. 4,B). We find that the time spent in association with DCs is 3-fold higher for CCR7 transgenic thymocytes compared with nontransgenic thymocytes. This reflects both an increase in the frequency of thymocyte-DC contacts, as well as a significant increase in the duration of contacts (Table I and Fig. 5,B), and is not accompanied by an increase in thymocyte speed (Figs. 4,B and 5 A, top panel). These results suggest that CCR7 expression on cortical thymocytes is sufficient to promote their association with DCs in the absence of other changes associated with positive selection.

Next, we asked if the loss of CCR7 in the background of a positively selecting TCR would result in fewer DC associations. We analyzed neonatal chimeras that were generated using bone marrow donor cells from CCR7^−/− mice (18) that were also P14 TCR and GFP transgenic (P14, CCR7ko → CD11cYFP). P14-expressing thymocytes lacking CCR7 spend 2-fold less time in association with DCs compared with their CCR7-expressing counterparts (P14 → CD11cYFP), corresponding to a decrease in the frequency of contacts (Table I). Although P14, CCR7ko thymocytes have an increased proportion of rapidly migrating cells compared with P14 thymocytes (Fig. 5,A), the reduced DC association is seen even when thymocytes of intermediate speed are compared (Fig. 4,B, highlighted region and Fig. 5 B, middle panel). These data suggest that up-regulation of CCR7 contributes to the increased DC associations, but not the increased speed, that normally accompany positive selection. Taken together, the correlation between CCR7 expression and DC contacts suggests that CCR7 up-regulation may contribute to increased DC scanning by thymocytes during positive selection via class I MHC.

Previous work has implicated chemokines in the long-range migration of mature thymocytes from the cortex to the medulla (5, 6). Here we provide evidence that CCR7 also plays a more localized role within the cortex to promote interactions between DCs and thymocytes that have begun to receive positive selection signals. We show that capillaries throughout the cortex colocalize with CCL21, and that thymic DCs form intimate associations with capillaries in the cortex. Increased association with cortical DCs occurs both for thymocytes that up-regulate CCR7 due to positive selection signals and for thymocytes expressing a constitutive CCR7 transgene. Conversely, the loss of CCR7 results in a reduction in thymocyte-DC association. Taken together, these data are consistent with a model in which CCL21 produced near vascular structures in the cortex attracts thymocytes that have initiated positive selection, thus promoting their interactions with DCs.

Interestingly, CCR7 up-regulation occurs to a greater extent on CD4⁺CD8⁺ thymocytes that are being selected via class I MHC compared with class II MHC (31). Although the current study focused on positive selection of thymocytes bearing class I-restricted TCRs, given that positive selection via class II MHC is also associated with CCR7 up-regulation relative to polyclonal or nonselected thymocytes, we would predict that similar, albeit less pronounced, DC associations should also occur during the positive selection of CD4 T cells. Future imaging studies using class II-specific TCR transgenic models should help to address this question.

At first glance the notion that CCR7 up-regulation on thymocytes could be involved both in promoting local interactions with DCs in the cortex and in promoting their long-range migration to the medulla seems paradoxical. However, positive selection is a prolonged process that occurs over a period of many hours and involves sets of cellular changes that occur with different kinetics (32, 33). This is reflected in expression changes for hundreds of genes, including a number with a potential role in controlling thymocyte motility (34, 35). These include genes encoding other chemokine receptors (CCR4 and CCR9) (29, 35), the axon guidance protein semaphorin 4A (34, 35), and the actin-severing protein gelsolin (34). It is possible that up-regulation of CCR7 represents a relatively early response to positive selection that, in the absence of other later changes, causes these thymocytes to respond to local sources of CCR7 ligands. At later stages of positive selection, additional changes may increase the overall motility of thymocytes or promote their detachment from cortical structures. The observation that loss of CCR7 reduces DC associations of class I-restricted thymocytes, but does not prevent the increase in motility associated with positive selection, is consistent with this view. In wild-type thymocytes, motility changes, together with CCR7 up-regulation, could promote rapid migration of mature thymocytes toward long-range sources of CCR7 ligands emanating from the medulla. Analysis of the kinetics of CCR7 up-regulation relative to other positive selection-induced changes and examination of how successive changes during positive selection affect thymocyte motility may shed light on these questions.

What are the implications of cortical thymocyte-DC interactions for TCR repertoire selection? Positive selection is thought to involve associations between thymocytes and cortical thymic epithelial cells, while thymic DCs are generally thought to mediate negative selection in the medulla (36, 37). Given the established role for thymic DCs in negative selection, our data suggest that a significant amount of screening for self-reactivity (negative selection) may occur in the cortex, overlapping in time and space with positive selection, and that positive selection-induced chemotaxis toward cortical DCs may enhance this screening. This is consistent with a recent report that CD4⁺CD8⁺ thymocytes from CCR7-deficient mice are partially spared from superantigen-mediated deletion (38). Moreover, in a TCR transgenic model for negative selection in which deletion occurs predominately in the cortex, apoptotic thymocytes are preferentially found in association with cortical DCs, and depletion of DCs partially abrogates negative selection (T. McCaughtry and K. Hogquist, unpublished data). Screening for negative selection in the cortex could help to eliminate thymocytes whose TCRs are reactive with ubiquitous self-Ags presented by DCs, as well as peripheral self-Ags carried into the thymus by migrating DC populations (37, 39, 40).

It is also possible that CCR7-driven interactions between cortical thymocytes and DCs could be related to positive selection. While expression of selecting MHC molecules on DCs alone is neither necessary nor sufficient to induce positive selection (41, 42), it is possible that MHC recognition during transient associations with DCs may help sustain the later phase of positive selection after the process is initiated by recognition of MHC on thymic epithelial cells. The observation that CCR7 up-regulation is more prominent for thymocytes undergoing positive selection via class I (31) suggests that such a function could be most relevant during the positive selection of CD8 T cells.

Finally, the location of thymocytes close to cortical DCs could facilitate phagocytosis of thymocytes that die as a result of positive or negative selection. While F4–80⁺ macrophages, rather than the cortical DCs described herein, appear to be the major phagocytic cell type in the mouse thymic cortex (14), we provide evidence that cortical DCs can also take up apoptotic cells. Functional overlap between DC and macrophage populations is a common theme in immunology, and it is also consistent with a recent report of a population in the human thymic cortex with both macrophage and DC-like properties (43).

More generally, we propose that the vasculature and its associated chemokines may organize specialized regions of the cortex by attracting thymocytes that have initiated positive selection to sites where repertoire selection may take place most efficiently. Such regions could serve to spatially segregate thymocytes that have begun the process of positive selection away from other thymocytes, thus reducing the effective volume for interactions with relevant support cells and promoting cellular interactions involved in repertoire selection. Cortical capillaries may also provide pathways for cell migration and/or could serve to bring cells in close proximity to oxygen and nutrients from the circulation. Understanding the role of these cortical structures awaits a more detailed analysis of their composition and the signaling and selection events that occur there.

We thank Kris Hogquist and Tom McCaughtry (University of Minnesota, Minneapolis, MN) for communication of unpublished data, Ying Fang Laio for help with manual scoring, Marie Le Borgne for help with imaging, Hector Nolla for assistance with flow cytometry, Michel Nussenzweig (Rockefeller University, New York, NY) for providing CD11cYFP transgenic mice, Nigel Killeen (University of California, San Francisco, CA) for providing CCR7 transgenic mice, Jason Cyster (University of California, San Francisco, CA) for providing CCL19Fc, plt/plt, and CCR7^−/− mice, Philippe Bousso (Pasteur Institute, Paris), B. J. Fowlkes (National Institutes of Health, Bethesda, MD), and members of the Robey Laboratory for comments on the manuscript, and Andrew Farr (University of Washington, Seattle) for helpful discussions.

The authors have no financial conflicts of interest.