Thymocytes bearing TCRs that recognize antigens expressed by medullary epithelial cells are eliminated by negative selection, but how thymocytes scan the medulla for negative selecting ligands, and the cellular dynamics of thymocyte-antigen presenting cell interactions during negative selection are not known. Here, we used two-photon imaging to address these questions. We found that medullary thymocytes move rapidly compared to cortical thymocytes, undergo confined migration, and make frequent, transient contacts with DCs. Auto-reactive thymocytes became further confined in their migration and migrated more slowly, making more extensive contacts with DCs that surrounded the confinement zones. Our data are consistent with a model in which some auto-reactive thymocytes do not undergo immediate arrest and cell death upon encountering their negative selecting ligands, but continue to scan potential antigen presenting cells in a local region of the medulla.
In the absence of negative selection, the average speed of medullary thymocytes (15 μm/min) is comparable to that of mature T cells in the lymph node14,24
. Previous observations showing that positive selection correlates with increased speed of cortical thymocytes, and that the fastest cortical thymocytes are migrating directionally toward the medulla18
, suggest that the increased speed of medullary thymocytes is part of the developmental program induced by positive selection. The high speed of medullary thymocytes contributes to their ability to contact multiple DCs, estimated at about 500 DCs total during the four days they spend in the medulla12
. However this may be an overestimate of the total number of individual DCs contacted, since the confined migration pattern of medullary thymocytes means that they may contact the same DC multiple times before migrating to another region.
In the presence of negative selecting ligand, medullary thymocytes moved more slowly, were more confined in their migration, and formed frequent and slightly prolonged contacts with DCs that surround the region of confined migration. Two-photon studies of lymph nodes have led to the notion of two phases of T cell priming; a first stage dominated by transient contacts with DC and a second stage characterized by stable interactions, with the first transient phase most prominent when TCR stimulation is suboptimal13, 32
. It is tempting to speculate that the transient interactions of auto-reactive thymocytes with DCs reported here may serve as phase one of negative selection, perhaps allowing thymocytes to confirm that they are auto-reactive before becoming committed to die. The slightly increased duration of contacts with DCs in the presence of the negative selecting ligand is consistent with this notion.
At first glance it seems surprising that thymocytes undergoing apoptosis were not a prominent feature of this model of negative selection. However, very few dying cells are detected in the thymus of adult mice, perhaps due to the efficient engulfment and removal of apoptotic thymocytes by macrophages33
. Moreover, it seems likely that auto-reactive OT1 thymocytes persist in the medulla for days before dying, based on the estimates of 4 day dwell time in the medulla12
, and the approximately 50% reduction in the number of medullary thymocytes in this model of negative selection. Thus hundreds of hours of cumulative imaging time for OT1 thymocytes in RIPmOVA hosts would be needed to record numerous examples of thymocyte death, rather than the ~24 hours of cumulative imaging time reported here. In addition, it is possible that the difficulty in observing thymocyte arrest and death in this negative selection model may reflect the low amount of self-antigen present in the AIRE-dependent form of negative selection; other forms of negative selection may proceed more rapidly. Indeed, the observation that superantigen-mediated deletion leads to the efficient removal of forbidden Vβ
s from the repertoire of mature CD4+
implies that negative selection occurs more rapidly in this setting.
It is worth considering these findings in light of what is known about the distribution of self-antigens in the medulla. Contact with five DCs per hour would likely be more than adequate to eliminate thymocytes that are specific for broadly expressed self-antigens. However, any given TRA may be expressed by only 0.5% to 2% of mTECs4
and thus may be quite restricted in its anatomical distribution in the medulla. While cross-presentation by DCs may broaden the distribution of TRAs somewhat, the observation that DCs are generally sessile in the thymus19
and this manuscript means that thymocytes must migrate in the vicinity of an mTEC that expresses the cognate self-antigen in order to be negatively selected. Moreover, it seems likely that the amount of TRA presentation may also be limiting, given that TRA expression on mTECs is low relative to the expression of the same antigen in peripheral tissues35
, and that an antigen may also have to make its way from an mTEC to a DC for cross-presentation. Thus, the requirement for a thymocyte to be able to respond to very low antigen concentration may be as important as the total number of DCs that can be scanned. In this regard, a scanning strategy in which thymocytes respond to a weak TCR signal by migrating in a confined area near the initial site of antigen contact could increase the chances that a thymocyte will re-encounter a self-antigen, thereby making the sampling of potential antigen-presenting cells more efficient.
It is unclear what anatomical features characterize the confinement zones of auto-reactive thymocytes. The average diameter of ~30 μm for a confinement zone is consistent with the possibility that confinement occurs due to the tethering of migrating auto-reactive thymocytes to a large antigen-presenting cell. Although there is evidence that DCs can cross-present antigen to OT1 T cells in this system7
, migration along the surface of a DC cannot account for the confined migration of auto-reactive thymocytes reported here, because confinement zones are found in between, not surrounding, the cell bodies of individual DCs. Medullary epithelial cells can also present directly to OT1 thymocytes in this system7
, and it is possible that one or a group of OVA-expressing mTECs could define a confinement zone. Another intriguing possibility is that confinement zones correspond to the sites of recent demise of an OVA-expressing mTEC. This possibility is suggested by recent evidence that mTECs turn over rapidly, and that AIRE expression promotes the death of the most mature mTECs10
. If antigens from dying mTECs become available for cross-presentation by nearby DCs, this could lead to the positioning of DCs presenting TRAs close to thymocytes that are deciding whether to undergo negative selection.
Perhaps the most surprising aspect to this study is the evidence that a substantial proportion of steady-state polyclonal medullary thymocytes display a migration signature indicative of negative selection. These data are consistent with the view that a substantial proportion of polyclonal thymocytes remain viable and motile for some time after they encounter negatively selecting ligands. The notion that thymocytes may encounter a negative selecting ligand and not be rapidly eliminated is also consistent with recent observations in a system for negative selection in the cortex28, 36
, and makes sense in light of recent observations that auto-reactive medullary thymocytes provide signals for the maturation of mTECs37–39
. The persistence and confined migration patterns of thymocytes undergoing negative selection may reflect their ongoing interactions with other cells as they help to construct and maintain medullary niches specialized in negative selection.