Directed Migration of Positively Selected Thymocytes Visualized in Real Time

Development of many vertebrate tissues involves long-range cell migrations. In most cases, these migrations have been inferred from analysis of single time points and the migration process has not been directly observed and quantitated in real time. In the mammalian adult thymus, immature CD4+CD8+ double-positive (DP) thymocytes are found in the outer cortex, whereas after T cell antigen receptor (TCR) repertoire selection, CD4+CD8– and CD4–CD8+ single-positive (SP) thymocytes are found in the central medulla. Here we have used two-photon laser-scanning microscopy and quantitative analysis of four-dimensional cell migration data to investigate the movement of thymocytes through the cortex in real time within intact thymic lobes. We show that prior to positive selection, cortical thymocytes exhibit random walk migration. In contrast, positive selection is correlated with the appearance of a thymocyte population displaying rapid, directed migration toward the medulla. These studies provide our first glimpse into the dynamics of developmentally programmed, long-range cell migration in the mammalian thymus.


Introduction
Although it is known that thymocytes relocalize from the cortex to the medulla after positive selection, the means by which this relocalization occurs is largely unknown [1][2][3]. For example, thymocytes may migrate by random walk throughout the cortex, with selected thymocytes being captured at the medulla by short-range guidance cues. Alternatively, thymocytes may migrate directionally across the cortex in response to long-range cues emanating from the medulla. Directional migration in response to such long-range cues may be induced by positive selection or may be a property of all cortical thymocytes. To address these and other questions, we undertook a real-time analysis of thymocyte migration in intact thymic lobes.

Results/Discussion
In order to track migrating thymocytes in situ, we generated chimeric mice in which a fraction of thymocytes express green fluorescent protein (GFP). We devised a protocol, based on a previously described method [4], in which adult bone marrow from GFP transgenic mice [5] is injected into newborn mice to generate partial GFP hematopoietic chimeras without the use of irradiation (see Materials and Methods). The GFP þ thymocytes generated in this way comprised approximately 1% of total thymocytes and showed expected developmental profiles by flow cytometric analysis ( Figure S1). At 4.5 to 5.5 wk of age, chimeric thymic lobes were harvested and imaged while being perfused with warmed oxygen-supplemented media. Imaging of intact lymph nodes under similar conditions revealed identical T cell and dendritic cell motility and behavior to that observed from intravital imaging of lymph nodes [6][7][8][9]. The objective was positioned above the center of the thymic lobe such that the movement of the stage (z direction) was perpendicular to the thymic capsule ( Figures S2 and S3). Regions of thymic cortex 104 3 104 3 40 lm were scanned at tissues depths between 80-200 lm below the surface of the capsule. The scan sequence was repeated every 37 s for up to 33 min to generate four-dimensional (4D) datasets (x, y, z, and time) (Videos S1 and S2). The datasets were analyzed with a combination of 4D cell-tracking software (Figure 1; Video S3) and mathematical treatment of the cell tracks.
Analysis of the motility rates of individual GFP cortical thymocytes ( Figure 2A) showed that the vast majority of cells (88% of total cells) moved at low motility rates between 3-8 lm/min (MR lo ) and exhibited peak instantaneous velocities of up to 9 lm/min ( Figure 2B). A small subset of imaged cells (approximately 7%) moved at motility rates that were 10 lm/ min or greater (MR hi ) and exhibited peak instantaneous velocities of up to 28 lm/min (Figure 2A and 2B; Video S4). The distribution of motility rates among cortical thymocytes was highly reproducible between samples ( Figure S4). The instantaneous velocities for each group showed no interconversion of motility rates throughout imaging ( Figure 2B), suggesting that these differing motility behaviors represented distinct cell populations.
Further analysis showed striking differences between MR lo and MR hi cells with regard to their morphology and with other aspects of their migratory behavior. For example, MR hi cells displayed a highly polarized morphology with a welldefined leading edge and uropod, and moved by a series of lurches followed by contraction ( Figure 2C; Video S5). In contrast, MR lo cells exhibited a more spherical morphology that lacked obvious polarization. Propulsion by MR lo cells involved only a modest protrusion of the cell's leading edge. These properties exclusively segregated with motility rates and remained constant over the entire imaging time. Also, whereas MR lo cells showed frequent pausing during the course of their trajectories, MR hi cells were never observed to pause (see Video S3; data not shown).
Examination of a cell's displacement from origin relative to time can provide additional insight into the migratory behavior of cells. Individual MR hi cells exhibit a linear relationship between displacement and time. In contrast, the displacement from origin for MR lo cells revealed numerous turns back toward cell origin ( Figure 2D). The turning behavior of thymocytes is of interest in part because it may reflect the interaction of thymocytes with other cells or structures in the tissue environment. For example, thymocytes engaged in dynamic contacts with MHC-bearing stromal cells during positive selection and T cells contacting antigenbearing dendritic cells in intact lymph nodes turn frequently and show little displacement over time [7][8][9][10]. We quantitated the degree of turning by computing the total length of a cell's trajectory divided by the absolute value of its displacement from origin. If a cell's migratory path shows little to no deviation from a straight line, this ratio will be close to one. An analysis of this directional index for each population showed a ratio of 5.9 for MR lo cells as compared to 1.6 for MR hi cells ( Figure 2E). Taken together with the frequent pausing observed for MR lo cells, but not for MR hi cells, these results suggest that the MR lo thymocytes interact with their environment more extensively than do MR hi thymocytes.
A major aim of this study was to determine whether the localization of mature thymocytes to the medulla involves directed inward migration across the cortex, and if so, whether directed migration is a property of all cortical thymocytes or only thymocytes that have been selected to mature. To examine this question, we used graphical techniques borrowed from diffusion mechanics to distinguish movement by random walk versus directed migration [11] (see Materials and Methods). Analysis of 1,250 MR lo cells showed that the relationship between displacement from origin with respect to time was consistent with random walk statistics ( Figure 2F). In contrast, a graph of the mean displacement from origin (as opposed to the square of the displacement) versus time for MR hi cells showed a linear relationship, indicative of directed migration. Thus, MR lo and MR hi thymocytes use distinct modes of migration as they move through the cortex.
The observation that MR hi cells moved by directed migration is consistent with the possibility that these cells are being directed to migrate toward the medulla. If this were the case, we would expect their trajectories to show a common orientation in the Àz direction (away from the capsule). To examine this question, we used two methods of statistical analysis (see Materials and Methods). First, vector analysis was performed in which the average displacement per cell in each direction of three-dimensional (3D) space for a fixed time interval of 3 min was calculated ( Figure 3A). If MR hi cell tracks were randomly oriented in the cortex, then the average displacement in each of the six directions (þx, Àx, þy, Ày, þz, and Àz) would be similar. On the other hand, if there were common directionality for MR hi cell trajectories, we should see an increase in average displacement values for the preferred direction. As shown in Figure 3A (and Figure S5), although the average displacements in the x and y directions were similar, there was greater displacement in both þz and Àz directions with the greatest bias in the þz direction (toward the capsule). Figure 3B provides a visual representation of track orientation shown along the x and z directions for both MR lo (top panel) and MR hi cells (lower panel).
To confirm and extend these results, we performed a step analysis on MR hi cells (see Materials and Methods) in which individual thymocytes were scored as showing net displacement in the positive or negative direction along each of three axes (x, y, and z). This analysis allowed us to correlate the tendency of thymocytes to migrate in a particular direction with their motility rates ( Figure 3C). As expected, thymocytes showed equal tendency to displace in the positive and negative directions along the x and y axes. In striking contrast, cells with motility rates greater than 13 lm/min were consistently scored as moving in the Àz direction (away from the capsule, toward the medulla). Interestingly, the vast majority of cells with motility rates in the range of 10À12 lm/ min were scored as moving in the þz direction (toward the capsule). We suspect that the population with intermediate motility rates is comprised of CD4 À CD8 À double-negative (DN) thymocytes based on published evidence for outward migration of DN thymocytes in the adult thymus [12,13]. CD4 þ CD8 þ double-positive (DP) thymocytes express clonally variable versions of the T cell antigen receptor (TCR). Following somatic V(D)J rearrangement and cell surface expression of the abTCR, cortical thymocytes test out their antigen receptors for their ability to bind self-peptide and MHC proteins expressed in the thymus. A small fraction of thymocytes expressing TCR with moderate avidity for selfpeptide MHC receive signals that allow them to differentiate into more mature medullary CD4 þ CD8 À or CD4 À CD8 þ thymocytes, a process known as positive selection [2,3,14]. The low frequency of cortical thymocytes with motility rates greater than 13 lm/min, together with their biased movement away from the capsule, led us to hypothesize that these cells might represent thymocytes that had successfully undergone positive selection.
To test this hypothesis, we generated chimeric mice in which a small fraction (approximately 1%) of thymocytes expressed both GFP and rearranged TCR transgenes that do or do not allow positive selection. As a positive-selecting TCR, we used the class I MHC-restricted P14 TCR transgene, which promotes the development of mature CD8 T cells in the H2 b (B6) background [15]. As a nonselecting TCR, we used the 5CC7 TCR transgene, which leads to neither positive nor negative selection in the B6 background [16]. Analysis of the motility rate distribution of cortical thymocytes expressing transgenic TCRs showed a striking correlation between positive selection and the frequency of thymocytes with high motility rates ( Figure 4A and Figure S6). A total of 34% of cortical thymocytes expressing the P14 TCR had motility rates greater than 13 lm/min compared to approximately 2% for wild-type thymocytes expressing diverse TCRs. In addition, thymocytes expressing the nonselecting 5CC7 TCR showed a nearly complete absence of cortical thymocytes moving at speeds greater than 13 lm/min. In fact, only three of the 1,275 5CC7 thymocytes that were analyzed exhibited motility rates greater than 13 lm/min. These values are significantly different from wild-type cortical thymocytes in which 50 of 1,670 thymocytes exhibited motility rates greater than 13 lm/min (p = 0.002). Moreover, although the majority of rapidly migrating wild-type thymocytes were migrating away from the capsule (44 of 50), all three of the rapidly migrating 5CC7 thymocytes were moving toward the capsule ( Figure 4).
Our studies show that positive selection leads to a rapid directional migration pattern and are consistent with earlier studies showing that activated CD4 þ CD8 þ cells migrate rapidly in vitro [17]. In an earlier study of thymocyte-stromal cell interactions in reaggregate thymic organ culture, we did not note a major difference in overall motility rates between positively selected and wild-type thymocytes [10]. This may be due to the fact that reaggregate thymic organ cultures lack the normal spatial distribution of chemokines and other guidance cues that are likely to control thymocyte migration patterns.
As in the case of MR hi cells of wild-type mice, P14 MR hi cells displayed a highly polarized morphology, and their trajectories showed very little turning, with no incidence of pausing ( Figure 4B and data not shown). In addition, P14 MR hi thymocytes displayed greater displacement in the positive and negative z directions (see Figure 4C), and step analysis showed that the vast majority of cells moving at rates greater than 13 lm/min moved away from the capsule ( Figure 4D). The motility rates and directionality of P14 thymocytes were highly reproducible between samples ( Figures S6 and S7).
When considering thymocytes with intermediate motility rates (10-12 lm/min), there were two notable differences between P14 and wild-type thymocytes ( Figure 4A and 4D). First, the frequency of these intermediate motility thymocytes was higher among P14 compared to wild-type cortical thymocytes ( Figure 4A). In addition, whereas the majority of wild-type thymocytes of intermediate motility migrated toward the capsule (see Figure 3C, right panel), this trend was less clear among P14 thymocytes ( Figure 4D, left panel). These differences could be explained by the proposition that thymocytes with intermediate motility rates consist of a mixture of outwardly migrating CD4 À CD8 À DN thymocytes and inwardly migrating positively selected CD4 þ CD8 þ thymocytes, with the DN subset predominating in wild-type samples. In P14 samples, the increase in the number of intermediate motility, inwardly migrating thymocytes could be attributed to an increase in the fraction of thymocytes undergoing positive selection.
Importantly, we observed directional migration of MR hi thymocytes in each dataset corresponding to a region of the cortex that extends from, approximately, 80 lm to 200 lm below the thymic capsule. This suggests that thymocyte migration is directed by guidance cues that extend over a large area of the cortex. Although the nature of these guidance cues is currently unknown, there are a number of chemokines expressed in the medulla whose corresponding receptors are upregulated during positive selection [18][19][20][21][22][23]. These include CCL19/CCL21, whose receptor, CCR7, is upregulated on activated CD4 þ CD8 þ thymocytes [19,24]. Moreover, gain and loss of function mutations have implicated CCR7 in the positioning of mature SP thymocytes to the medulla [24,25]. The contribution of CCR7 and other chemokine receptors in controlling the thymocyte migration patterns described here will be an important area for future investigation.
A cortical thymocyte must travel a distance of hundreds of microns in order to reach the medulla. Based on the average distance from the capsule to the medulla in the adult mouse thymus, and the speed and directionality reported here, we estimate that a typical MR hi thymocyte that we image in the cortex could arrive at the medulla in 1 to 2 h. This short time period for migration to the medulla is in contrast to the estimates of 2-3 d for a CD4 þ CD8 þ thymocyte to complete positive selection [26,27]. Moreover, we have previously shown that thymocytes frequently pause and turn during MHC-driven contacts with stromal cells [10], behaviors that differ strikingly from the behavior of MR hi thymocytes described here. Based on these considerations, we suspect  Figure 2A) from wild-type (WT) thymocytes (red) were overlaid for comparison. P14 cells moving at motility rates greater than 13 lm/min were 34% of total imaged thymocytes (Videos S6 and S7) compared to approximately 1% of wild-type cortical thymocytes. Analysis of 5CC7 thymocytes showed nearly complete absence of cells moving at motility rates greater than 13 lm/min, a value that differed significantly (p = 0.002) from wild-type cortical thymocytes. that thymocytes moving at rates greater than 13 lm/min represent cells at a relatively late stage in the positiveselection process and that migration from the cortex to the medulla may not require continuing MHC engagement. In contrast, the MR lo cells are likely to include thymocytes that are actively engaging thymic stromal cells and receiving MHC-driven TCR signals. Future analysis of the signaling events and migratory patterns of these slowly migrating thymocytes may provide further insights into the process of positive selection in the thymus.

Materials and Methods
Generation of GFP chimeric mice. Mice expressing a GFP transgene driven by the ubiquitin promoter [5] were used as bone marrow donors for the generation of GFP hematopoietic stem cell chimeras using a modification of a previously described procedure [4]. Whole bone marrow from a single adult GFP þ mouse was aseptically harvested and resuspended into a single cell suspension in sterile Hanks' Balanced Salt Solution (Mediatech Cellgro). A total of 2-3 3 10 6 bone marrow cells were injected into newborn mice (C57Bl/6) in a volume of 70 ll. The first injection was done at 12-24 h after birth and repeated every 2-3 d for a total of four injections. Resulting chimeric mice expressed GFP in 1-2% of their thymocytes. P14 TCR transgenic mice [15] on a Rag2 À/À B6 background were obtained from Taconic. P14 þ/þ Rag2 À/À mice were crossed with UBI-GFP transgenic mice to generate P14 þ/-Rag2 þ/À GFP þ/À mice, and these mice were then intercrossed to generate P14 þ/þ or þ/À Rag2 À/À GFP þ/þ or þ/À mice. 5CC7 TCR transgenic mice [16] on a Rag2 À/À B10 background were obtained from Taconic and were crossed once with UBI-GFP transgenic mice to generate 5CC7 þ/À GFP þ/À Rag2 þ/À mice. Bone marrow from adult double transgenic mice was used to generate chimeric mice as described above.
Two-photon imaging of intact thymic lobes. Thymi from 4.5-5.5 wk-old GFP chimeric mice were quickly harvested, lobes were separated, and the dorsal face of the lobe was adhered to 22 3 22 mm cover glass with single drop of Vetbond tissue adhesive (see Figure S2). Cover slip with lobe was immediately placed into a 60 3 15 mm polystyrene Petri dish containing Dulbecco's modified Eagle's medium (DMEM) without phenol red (Mediatech Cellgro). Petri dish was placed into a heated ring, and the sample was perfused with warmed media bubbled with a blend of 95% O 2 and 5% CO 2 . Sample was maintained under perfusion and held at 36.5 8C to 37.5 8C throughout imaging. Thymic lobes maintained under these conditions for up to 6 h showed no changes in cell motility and no indications of tissue deterioration. Imaging was performed as previously described [8] using an upright Zeiss NLO 510 microscope equipped with a MaiTai Ti:Sapphire laser (Spectra-Physics). For each dataset, 20-33 min of imaging was performed with the objective oriented over the top center of the thymic lobe (corresponding to the ventral side of the organ; see Figures S2 and S3). A total of 20 optical slices were acquired at 2-lm step intervals with a total acquisition time of 36.7 s/z stack. Using Imaris Bitplane software, z stacks (104 3 104 3 40 lm in dimension) were processed into 3D images and reiterated through time to generate a 3D movie of thymocyte migration. In most experiments, data were acquired as ''blocks'' of stacked movies; a first movie was made at a maximum depth below the capsule with a second movie acquired immediately above the bottom movie (À180 to À80 lm; see Figure S3B). Analysis of HNEstained thymic tissue sections indicated that the area imaged invariably corresponded to cortex (data not shown).
4D data analysis. The 4D cell tracking was performed on 75-300 cells per movie using Imaris Bitplane software which identifies the x, y, and z coordinates for each cell at each given time point. These statistics were exported into an Excel spreadsheet for analysis. Average motility rates (MR) were computed as the total length of migratory path divided by the total time of tracking. The extent to which a cell's migratory path deviated from a straight line was quantitated as the total length of a cell's trajectory divided by the total displacement from origin. For the determination of movement by random walk for MR lo cells, the mean-square displacement from origin was shown to be proportional to time. This relationship is given by hr 2 i = 6Dt, where r = the displacement from origin, h i denotes the average r over numerous events at time t, and 6D is the motility coefficient [9], which characterizes the spread of cells in three dimensions. Movement by directed migration for MR hi cells was indicated by a linear relationship between mean displacement from origin (as opposed to its square) and time when computed over many events. The correlation coefficient R 2 for the best-fit line was computed in Excel using the least squares method. R 2 at p = 0.001 is statistically significant at values of 0.801 or greater. For displacement analyses of MR hi cells, the average displacement per cell in each direction was calculated from 3 min of tracking. For step analyses, thymocytes were grouped according to their average motility rates and scored as showing net displacement in the positive or negative directions along each of three axes (x, y, and z). The percentage of cells in each motility rate category which moved in the positive and negative directions was then graphed as a function of average motility rate. The computation of statistical significance between the frequencies of 5CC7 TCR thymocytes with high motility rates as compared to wild-type thymocytes was done in Excel using a paired t test.
Supporting Information Figure S1. Developmental Profiles of GFPþ Thymocytes from Chimeric Mice Representative profiles obtained by flow cytometric analysis of P14 and 5CC7 chimeric thymii. As expected, GFP þ -gated P14 thymocytes (top row) showed high levels of TCR within the CD4 þ CD8 þ population and a high percentage of CD8 þ SP thymocytes, indicating a high frequency of positive selection. As expected for expression of 5CC7 in a nonselecting host (bottom row), thymocytes remained arrested at the CD4 þ CD8 þ stage of development and fail to upregulate TCR. Found at DOI: 10.1371/journal.pbio.0030160.sg001 (128 KB TIF). Figure S2. Orientation of Imaging Relative to Thymic Lobes In Vivo Thymic lobes are depicted in their normal position relative to the heart. Thymic lobes were surgically removed and separated, and then the dorsal side (side facing the heart) of thymic lobe was adhered to glass cover slip. Imaging (see Figure S3A and S3B) was performed with the objective positioned over the center of the ventral side of lobe. Found at DOI: 10.1371/journal.pbio.0030160.sg001 (223 KB TIF). Figure S3. Two-Photon Imaging of Thymocyte Migration in Intact Thymic Lobes (A) Explanted GFP chimeric thymic lobe was placed in oxygenperfused media and maintained at 37 8C throughout experiment. Objective was placed directly over the top of lobe and a total of 20 optical slices at 2-lm step intervals were acquired, which generated z stacks of 104 x 104 x 40 lm in the x, y, and z directions. The z stack acquisition was repeated every 37 s for 20-33 min. Stacks were rendered into 3D images and processed through time to yield 4D datasets (see Videos S1-S8). (B) In most cases, a stack of movies was generated to increase the effective area of imaging. A bottom movie was generated by imaging starting at À160 to À200 lm below capsule and then a second movie was generated starting 2 lm above the bottom movie.     Video S1. GFP Thymocytes within an Intact Thymic Lobe A representative 3D image of GFP thymocytes within an intact thymic lobe. Image is rendered from one z stack at a single time point and is shown in a 3608 rotation. Image size is 164 3 164 3 40 lm. Image was recorded approximately 140 lm below the thymic capsule. Corresponds to Figure 1. Video S5. MR hi Cell Propulsion Is Associated with Polarized Morphology Time-lapse image of GFP thymocytes cropped to approximately 40 3 40 3 40 lm in the x, y, and z directions. Note the polarized morphology and dramatic shape changes of the MR hi cell as it crawls from bottom to upper left corner. Video shown was generated from 5 min of imaging and is played at six frames per second. Corresponds to Figure 2C.
Video S6. P14 TCR Transgenic GFP Thymocytes in an Intact Thymic Lobe Time-lapse image of P14 TCR transgenic GFP thymocytes in an intact thymic lobe. The P14 TCR induces positive selection in this system. Note that a high proportion of thymocytes migrate rapidly and in straight trajectories compared to wild-type GFP thymocytes (Videos S2 and S3). Corresponds to Figure 4B. Time-lapse image of 5CC7 TCR transgenic GFP thymocytes in an intact thymic lobe. The 5CC7 TCR is nonselecting in this system. Note the almost complete absence of rapidly migrating thymocytes. Found at DOI: 10.1371/journal.pbio.0030160.sv008 (1.1 MB ZIP).