Sphingosine-1-phosphate receptor 2 is critical for follicular helper T cell retention in germinal centers

First, we tested for functional expression of S1PR2 in CXCR5^hiPD-1^hi Tfh cells by performing transwell migration analysis (Fig. 1 A). Migration of these cells toward CXCL13 and CXCL12 (Ansel et al., 1999) was suppressed by S1P. This suppression by S1P was reversed by treatment with the S1PR2 antagonist JTE-013, suggesting that the suppression was mediated by S1PR2. These results are consistent with the previously described function of S1PR2 that inhibits Rac-mediated chemotaxis by Rho activation (Skoura and Hla, 2009). In contrast, S1P rather induced migration of CXCR5⁻CD4⁺ T cells, which was most likely mediated by G_i signaling-coupled S1P receptors, particularly S1PR1 (Matloubian et al., 2004). JTE-013 did not affect S1P- or CXCL12-induced migration of CXCR5⁻CD4⁺ T cells, suggesting that S1PR2 expression is minimal in these cells.

The modest S1pr2 expression in Tfh cells, together with evidence that GC-associated Tfh cells differ phenotypically from non-GC–associated Tfh cells, led us to speculate that S1PR2 expression is enriched in a subfraction of CXCR5^hiPD-1^hi Tfh cells localized in GCs. To test this hypothesis, we generated a reporter mouse strain by gene targeting for detecting S1pr2 expression in individual cells. A large portion of S1pr2 gene protein coding region was replaced with the yellow fluorescent protein (Venus) gene (Fig. S1 A). Flow cytometry of Peyer’s patch (PP) cells from S1pr2^Venus/+ (S1pr2^V/+) mice showed that GC B cells expressed uniformly high levels of Venus, and non-GC B cells were Venus negative (Fig. 1 B and Fig. S1 B). These observations are consistent with the previous reports (Cattoretti et al., 2009; Green et al., 2011). Within CD4⁺ T cell populations, the highest Venus signal was detected in Tfh cells, some of which expressed comparable levels of Venus as GC B cells. CXCR5^intPD-1^int Th cells expressed intermediate levels of Venus, and naive CD4⁺ T cells were Venus-negative (Fig. 1 B; and Fig. S1, B and C). Similar results were obtained for transferred S1pr2^V/+ OVA-specific TCR transgenic (OT-II) CD4⁺ T cells (unpublished data). Kinetics of Venus^hi OT-II Tfh cell development was slower than that of entire OT-II Tfh cell development, and was similar to that of GC B cell development (Fig. 1 C and Fig. S1 D). It has been recently found that FoxP3⁺ regulatory T cells also contain CXCR5^hiPD-1^hi cells, and these cells are called follicular regulatory T (Tfr) cells (Ramiscal and Vinuesa, 2013). Analysis of PP cells from S1pr2^V/+ mice that also carry the Foxp3 reporter transgene (hCD2) showed that Tfr cells also expressed S1pr2 (Fig. 1 D and Fig. S1 E). Tfr cells expressing S1pr2 as highly as GC B cells were 22 ± 1.8% of Tfr cells, which was significantly lower than the percentage of S1pr2-high Tfh cells in Tfh cells (37 ± 2.1%, mean ± SEM, n = 5, P = 0.0005). These data suggest that Tfh cells and Tfr cells express various levels of S1PR2, and only subpopulations of these cells express S1PR2 as highly as GC B cells.

To examine a role for S1PR2 in Tfh cell localization, histological analysis was performed. S1pr2^V/+ OT-II T cells were transferred to congenic recipient mice, and draining LNs were analyzed after immunization with (4-hydroxy-3-nitrophenyl)acetyl (NP) coupled to OVA (NP-OVA). 7 d after immunization, S1pr2^V/+ OT-II T cells expressing histologically detectable amounts of Venus were found mainly in the GC (Fig. 2 A). Venus expression levels in most of S1pr2^V/+ OT-II T cells in the FM and in the T cell zone were below the detection limit. In contrast, S1pr2^V/V OT-II T cells expressing detectable amounts of Venus were not confined in the GC, and the majority of them were found in the FM but not in the T cell zone (Fig. 2 A). Interpretation of this observation requires caution because twice the number of the Venus transgenes can lead to an overestimation of the effect of S1PR2 deficiency. More importantly, however, the number of S1pr2^V/V OT-II T cells in the GC was significantly reduced compared with S1pr2^+/+ OT-II T cells and S1pr2^V/+ OT-II T cells (Fig. 2, A and B) even though the formation of Tfh cells or expression of ICOS, Bcl6, and Il21 was not significantly changed by S1PR2 deficiency (Fig. 2, C and D; and Fig. S2). Consistent with the previous report, S1PR2 deficiency augmented numbers of GC B cells in mucosal lymphoid organs, which contain chronic GCs (Green et al., 2011). However, Tfh cell numbers were not significantly changed by S1PR2 deficiency even in the mucosal lymphoid organs (unpublished data). These results strongly suggest that the S1PR2 expression does not affect the formation of Tfh cells but is important for their localization in the GC.

To better understand the role for S1PR2 in Tfh cell dynamics in GC-containing follicles, we performed real-time two-photon microscopy. Avian-lysozyme–specific B cell receptor expressing Hy10 B cells harboring the cyan fluorescent protein (CFP) transgene and S1pr2^+/+ or S1pr2^V/V OT-II T cells harboring the GFP transgene were transferred to recipient mice, and explanted LNs were observed 7 or 8 d after immunization with hen egg lysozyme (HEL) conjugated with OVA (HEL-OVA) in alum. GFP⁺ OT-II T cells were visualized in follicles that contained GCs formed by CFP⁺ Hy10 B cells (Fig. 3 A; and Videos 1 and 2). Centroids of OT-II T cells that came into the GC-FM interface zone were tracked to determine which side of the interface zone they migrated from and to (Video 3). A majority of S1pr2^+/+OT-II T cells that came from the GC to the interface zone were observed to return to the GC, consistent with the previous report (Fig. 3 B; Qi et al., 2008). In contrast, S1pr2^V/V OT-II T cells that came from the GC to the interface zone failed to show the preference for returning to the GC (Fig. 3 B). OT-II T cells that came from the FM to the interface zone showed a clear preference for returning to the FM, regardless of their genotype (Fig. 3 C). These results, together with the flow cytometry data (Fig. 1) and the histological data (Fig. 2), suggest that S1PR2 is important for retaining the S1PR2^hi subpopulation of Tfh cells in the GC.

To further characterize the Tfh cell subpopulations expressing different levels of S1pr2, we sorted Venus^hi Tfh cells, Venus^lo Tfh cells, PD-1^int Th cells, PD-1^lo Th cells, and naive CD4⁺ T cells (Fig. S3 A) and compared their mRNA expression profiles by microarray analysis (Fig. 4 A). Although Venus^hi Tfh cells and Venus^lo Tfh cells showed generally similar gene expression profiles, there were several notable differences. Bcl6 expression was higher in Venus^hi Tfh cells than in Venus^lo Tfh cells and in PD-1^int Th cells. Ccr7 expression was low but still detectable in Venus^lo Tfh cells, whereas it was virtually abrogated in Venus^hi Tfh cells. Expression of several other G_i-coupled receptor genes, Ccr6, Gpr18, Gpr183, and S1pr1, was lower in Venus^hi Tfh cells than in Venus^lo Tfh cells, which might redundantly contribute to the segregation of Venus^hi Tfh cells in the GC and Venus^lo Tfh cells in the FM. Microarray data showed that Venus^hi Tfh cells expressed slightly more transcripts of Il4 (1.41-fold) and Il21 (1.23-fold) than Venus^lo Tfh cells. To more clearly address these points, we crossed the S1pr2^Venus mice with two different reporter mouse strains to detect Il4 or Il21 expression. Consistent with the previous reports, expression of the Il4 and Il21 reporters (human CD2) was detected selectively in Tfh cells (Reinhardt et al., 2009; Harada et al., 2012; Lüthje et al., 2012; Fig. S3 B). Within Tfh cells, the percentage of cells expressing the Il4 and Il21 reporters was higher in the Venus^hi population (Fig. 4, B and C; Fig. S3 C). These results suggest that S1PR2^hi Tfh cells most abundantly express the important molecules for Tfh cell function.

Because S1PR2 deficiency reduced the number of Tfh cells in GCs by half (Fig. 2 B), we sought to investigate its impact on GC responses. To eliminate the effects of S1PR2 deficiency in B cells (Green et al., 2011), S1pr2^V/V or control CD4⁺ T cells were transferred to Cd28^−/− mice, which have a much reduced number of Tfh cells (Linterman et al., 2009), and analyzed after immunization. As shown in Fig. 5 A, S1pr2^V/V and control CD4⁺ T cells could similarly support GC B cells. Histological analysis showed that the number of CD4⁺ T cells in GCs was significantly reduced in S1pr2^V/V CD4⁺ T cell–transferred mice compared with control CD4⁺ T cell–transferred mice, consistent with the data shown in Fig. 2 B (Fig. 5 B and Fig. S4 A). The area of a GC section was not significantly different between two groups (Fig. S4 B). The ratio of CXCR5^hi Tfh/GC B cell numbers was not affected by S1PR2 deficiency in T cells, indicating that total Tfh cell formation was unaffected (Fig. S4 C). We also observed similar GC formation between two groups of sublethally irradiated Rag1-deficient mice, one group reconstituted with S1pr2^+/+ μMT BM cells mixed with Tcrα^−/− BM cells and the other reconstituted with S1pr2^V/V μMT BM cells mixed with Tcrα^−/− BM cells, after immunization with NP–chicken gamma globulin (CGG) in alum i.p. (unpublished data). In addition, both chimera types produced similar amounts of anti-NP1 and NP24 IgG1, and the ratio of NP1/NP24 was comparable (Fig. 5 C and Fig. S4 D). These results suggest that, contrary to our expectations, the partial decrease of GC Tfh cells caused by S1PR2 deficiency alone may not significantly affect GC B cell maintenance at least in these experimental settings. Because previous studies showed that T cell–specific CXCR5 deficiency only mildly reduced GC responses (Junt et al., 2005; Arnold et al., 2007; Haynes et al., 2007), we next examined if deficiency of both CXCR5 and S1PR2 in Th cells would further suppress GC responses. WT, Cxcr5^−/−, or Cxcr5^−/− S1pr2^V/V CD4⁺ T cells were transferred to Cd28^−/− mice, and the percentage of GC B cells was examined after immunization. In agreement with the previous studies, GC B cell formation supported by Cxcr5^−/− CD4⁺ T cells was partially impaired compared with that supported by WT CD4⁺ T cells (Arnold et al., 2007; Haynes et al., 2007). Strikingly, GC responses were further reduced when the transferred CD4⁺ T cells lacked both CXCR5 and S1PR2 (Fig. 5 D). Percentages of activated cells in transferred CD4⁺ T cells of the individual genotypes were not significantly different, suggesting that priming of S1PR2 and CXCR5 double-deficient CD4⁺ T cells occurred normally (Fig. S4 E). Histological analysis confirmed that GCs supported by CXCR5 and S1PR2 double-deficient T cells were hypoplastic and scarcely even contained T cells compared with GCs supported by CXCR5 single-deficient T cells (unpublished data). These results strongly suggest that S1PR2 in Tfh cells has a cooperative role with CXCR5 in supporting GC responses.

To quantitatively determine the effect of S1PR2 and CXCR5 double deficiency on T cell localization in well-formed GCs, we cotransferred CD45.1 WT CD4⁺ T cells with CD45.2 WT, Cxcr5^−/− S1pr2^V/+, or Cxcr5^−/− S1pr2^V/V CD4⁺ T cells to Cd28^−/− mice, and immunized the mice. Histological analysis revealed that the number of CD45.2 CD4⁺ T cells in GCs relative to the number of CD45.1 CD4⁺ T cells in GCs was decreased when transferred CD45.2 CD4⁺ T cells were CXCR5-deficient, and was even further reduced when transferred CD45.2 CD4⁺ T cells were S1PR2 and CXCR5 double-deficient (Fig. 5 E, and Fig. S4 F). These results, together with the results shown in Figs. 2 and 3, strongly suggest that S1PR2 and CXCR5 in Tfh cells act additively and are required for their localization in GCs.

In essence, our data indicate that GC Tfh cells are retained within GCs by means of their S1PR2 expression. S1PR2 has been previously shown to contribute to accumulation of GC B cells in the central region of the follicle (Green et al., 2011; Wang et al., 2011). This is most likely mediated by negative concentration gradients of S1P from the perimeter to the center of the follicle, established through the actions of S1P degrading enzymes (Green and Cyster, 2012). In addition to retention in GCs, it is possible that S1PR2 may promote Tfh cell recruitment to GCs if they start to express S1PR2 while they are outside of GCs. Our analysis also shows that S1PR2 mediates CXCR5-independent GC association of T cells. We speculate that S1P concentration may be lower in the GC than in the T cell zone, and that CXCR5-deficient Th cells that have down-regulated CCR7 enter follicles at a relatively low frequency, after which some of them may be retained in GCs due to the S1PR2-mediated inhibition of migration back to the T cell zone.

A very recent study using the in vivo photoconversion technique has shown that although Tfh cells become concentrated in GCs over time, they can exchange between GCs within a responding LN (Shulman et al., 2013). After 20 h, about one third of photoconverted Tfh cells from a single GC were found in neighboring follicles and other GCs. This indicates a residence half-time in a given GC exceeding 1 d, considerably longer than the residence half-time of naive T cells in an entire LN (Lo et al., 2005). Given the small size of the GC structure, these data are consistent with strong confinement of most GC Tfh cells to the GC. It will be important in future work to assess if GC Tfh cells down-modulate S1PR2 expression over the course of immune response or in response to antigen boost to increase the extent of exchange between GCs.

Our data have demonstrated that GC formation is severely impaired when T cells lack not only CXCR5 but also S1PR2, suggesting that S1PR2-mediated Tfh cell retention in GCs contributes to the maintenance of GC responses. At this point, however, it is not excluded that possible additional roles of S1PR2 signaling in Tfh cells may also contribute to GC responses, as S1PR2 signaling in GC B cells regulates not only their localization but also their survival.

The present study demonstrates the cooperative interplay between CXCR5 and S1PR2 for Tfh cells to promote GC responses, which exemplifies that a G_i-coupled receptor and a G_12/13-coupled receptor can work redundantly for cells to exert their function in vivo. Nonetheless, it is still possible that S1PR2 in Tfh cells has nonredundant roles for GC responses. As S1PR2 deficiency in T cells is likely to reduce the number of GC Tfr cells, which may counteract the effect of the partial reduction of GC Tfh cells (Ramiscal and Vinuesa, 2013), it will be important to test the outcome of Tfh cell–specific S1PR2 deficiency. Future studies will also need to test the impact of S1PR2 deficiency in T cells on long-term antibody responses against viruses and on antibody diversification to maintain balanced microbiota in the gut.

To generate the S1pr2^Venus mice, the first 498 bp of the S1pr2 protein coding region was replaced with the Venus gene. A loxP-flanked PGK-Neo cassette was inserted before exon 2, a 7.1-kb fragment was used as the 5′ homology region, and a 2-kb fragment was used as the 3′ homology region. C57BL/6J × C57BL/6N hybrid ES cells were transfected, cultured, and selected. Then, correctly targeted clones were aggregated with BALB/c morulas. Chimeric male offspring were mated with C57BL/6J female mice, and their offspring were mated with CAG-Cre transgenic mice to excise the PGK-Neo cassette. Approximately 40% of S1pr2^V/V mice at the age of 3–5 wk were found dead, often in seizure-like postures as previously described for the other S1PR2-deficient mouse lines (MacLennan et al., 2001; Du et al., 2010). C57BL/6J mice were purchased from CLEA Japan. Rag1-deficient mice were bred and maintained in CLEA Japan. β-actin–CFP (The Jackson Laboratory), CD28-deficient (The Jackson Laboratory), CD45.1 congenic (The Jackson Laboratory), CXCR5-deficient (The Jackson Laboratory), Foxp3^hCD2 (Miyao et al., 2012), HEL-specific Ig–knock-in Hy10 (previously named VDJ9/κ5; Allen et al., 2007), IL-4/hCD2 transgenic, IL-21/hCD2 transgenic (Harada et al., 2012), μMT (The Jackson Laboratory), OT-II (The Jackson Laboratory), TCR-α–deficient (The Jackson Laboratory; Mombaerts et al., 1992), and UBC-GFP (The Jackson Laboratory) mice were bred and maintained in specific pathogen-free conditions at the animal facility of RIKEN Center of Integrative Medical Sciences. For BM reconstitution, BM cells were injected i.v. to 6 Gray-irradiated Rag1-deficient mice, and the mice were immunized 3 mo after reconstitution. All animal studies were performed in accordance with the guidelines of the RIKEN Animal Research Committee and were conducted under protocols approved by the Animal Experiments Committee of RIKEN Yokohama Institute.

Splenocytes, LN cells, and PP cells were harvested by mechanical disruption with slides (Matsunami Glass) or on 40-µm nylon cell strainers (BD) in DMEM containing 2% FBS. BM cells were flushed from femurs and tibiae in DMEM containing 2% FBS. After removal of red blood cells if necessary, non-CD4⁺ cells were labeled with biotinylated anti-B220 (RA3-6B2), anti-CD8a (53–6.7), anti-CD11b (M1/70.15), anti-CD11c (HL3), anti–Gr-1 (RB6-8C5), and anti-NK1.1 (PK136), followed by Streptavidin MicroBeads (Miltenyi Biotec). Non–B cells were directly labeled with CD43 MicroBeads (Miltenyi Biotec). Samples were magnetically depleted of labeled cells with the autoMACS cell separator (Miltenyi Biotec) or using CS columns (Miltenyi Biotec), and the purity of CD4⁺ T cells and B220⁺ B cells was typically >80% and >95%, respectively. For two-photon imaging, purified B cells were labeled with 20 µM 5-(and-6)-(((4-chloromethyl)benzoyl)amino)tetramethylrhodamine (CMTMR; Invitrogen) in DMEM containing 2% FBS for 20 min at 37°C. 1–3 d before imaging, 1.4–3.0 × 10⁷ CMTMR-labeled polyclonal B cells were transferred i.v. In most experiments, 1–5 × 10⁵ Ag-specific T cells and B cells were transferred. The next day, the mice were immunized s.c. in the footpad, base of the tail, flank, and scruff with 10–15 µg NP(16)-OVA (Bioresearch Technologies), HEL-OVA, CGG (EMD Millipore), or NP-CGG mixed in alum (Imject alum; Thermo Fisher Scientific) or CFA (Sigma-Aldrich) per site. For the time course experiment, S1pr2^V/+ OT-II T cells and Hy10 B cells (2 × 10⁵ each per head) were cotransferred into recipient mice which were then s.c. immunized with HEL-OVA in CFA 1 d after cell transfer. The results about flow cytometry patterns of S1pr2-Venus versus CXCR5/PD-1 are very similar between CFA immunization and alum immunization.

Explanted LNs were imaged as previously described (Kitano et al., 2011). Multiphoton excitation was provided by a Mai Tai HP Ti:Sapphire laser (Spectra Physics) tuned to 900 nm, and images were acquired with a 20×/1.00 NA water immersion objective (Leica). Clusters of CFP⁺ Hy10 B cells in which transferred CMTMR-labeled polyclonal B cells were apparently sparse compared with the region outside the clusters were used as the GC region for subsequent analysis. Areas within 10 µm of the GC surface reconstructed from CFP⁺ B cell clusters were defined as the GC-FM interface zone. T cell tracking and the creation of GC surface were performed using Imaris 64× 7.2.3 (Bitplane) tracking and surface command. Tracks (>8 min) of T cells that moved into and then left the GC-FM interface zone were used for quantitative analysis.

The cells were stained anti-B220 (RA3-6B2, APC/Cy7), anti-Bcl6 (K112-91, PE), anti-CD4 (RM4-5, PE/Cy7 and PerCP-Cy5.5), anti-CD38 (90, APC), anti-CD44 (IM7, APC/Cy7 and PE/Cy7), anti-CD45.1 (A20; PerCP-Cy5.5, Pacific blue and APC/Cy7), anti-CD45.2 (104, Pacific blue and PE), anti-CD62L (MEL-14, APC and PerCP-Cy5.5), anti-CD95 (Jo2, PE/Cy7), anti–human CD2 (RPA-2.10, PE), anti-CXCR5 (2G8, APC), anti-IgD (11-26c, eFluor450), anti-ICOS (15F9, PE), anti–Ly-77 (GL7, biotin), anti–PD-1 (J43, PE), and streptavidin reagents (Pacific blue, APC, PE and PerCP-Cy5.5) and analyzed on a FACSCanto II (BD). Bcl6 staining was performed as previously described (Kitano et al., 2011). Antibodies were purchased from BD, BioLegend, or eBioscience. The data were analyzed with FlowJo software (Tree Star).

Draining LNs and spleens were fixed with 4% paraformaldehyde, washed with PBS, and then placed in Tissue-Tek OCT compound (Sakura) and snap-frozen in dry ice and ethanol. Cryosections 8 µm in thickness were affixed to MAS-GP–coated slides (Matsunami Glass) and stained at room temperature for 2 h or at 4°C for overnight with anti-IgD (11-26c.2a, FITC and PE), anti-CD4 (RM4-5, PE), anti-CD45.1 (A20, APC), anti-CD45.2 (104, APC), and rabbit anti–Bcl-6 (N-3; Santa Cruz Biotechnology, Inc.), which was detected by FITC-conjugated goat anti–rabbit IgG (SouthernBiotech) or by Alexa Fluor 488–conjugated goat anti–rabbit IgG (Invitrogen). Some samples were dehydrated with ethanol after sectioning instead of 4% paraformaldehyde fixation. Venus expression was detected by goat anti-GFP antibody (Rockland), biotinylated donkey anti–goat IgG antibody (Jackson ImmunoResearch Laboratories, Inc.), and TSA kit with HRP-streptavidin and Alexa Fluor 488 or Alexa Fluor 546 tyramide (Invitrogen). Images were acquired on BZ-9000 (Keyence), and processed with Photoshop (CS5.1; Adobe). ImageJ (National Institutes of Health) was used for the measurement of follicle and GC area for normalization of Tfh cell counts.

Total RNA was isolated from FACSAria II (BD)–sorted T cells using TRIzol reagent (Invitrogen). Biotinylated cDNA was synthesized using the Ovation RNA Amplification System V2 (Nugen) and FL-Ovation cDNA Biotin Module V2 (Nugen), and then hybridized with Mouse Genome 430 2.0 Arrays (Affymetrix). Expression values for each probe set were calculated with the GC-RMA method in the GeneSpring GX 7.3 software package (Agilent Technologies).

Splenic CD4⁺ T cells were prepared from C57BL/6 mice 10–12 d after sheep red blood cell immunization, and resensitized for 40 min at 37°C before adding to 5 µm transwells. 1 µg/ml CXCL13 (R&D Systems), 250 ng/ml CXCL12 (PeproTech), 10 µM JTE-013 (Tocris Bioscience), and S1P (Sigma-Aldrich) was added to the lower chamber as indicated in the figure, and analyzed as previously described (Green et al., 2011). In our experiments, 20–35% of input CXCR5^hi T cells migrated in response to CXCL13, and 4–6% of input CXCR5^hi T cells migrated without chemokines. To calculate the chemotactic index, the number of cells that migrated in response to chemokines and S1P was divided by the number of cells that migrated without chemokines or S1P.

96-well plates (Thermo Fisher Scientific) were coated overnight at 4°C with 2 µg/ml NP1-BSA or NP24-BSA diluted with borate-buffered saline. After washing, PBS containing 0.5% BSA and 0.05% Tween-20 was added to the plates to prevent nonspecific binding. Then, serial dilutions of sera and anti-NP monoclonal antibody standard (C6) were applied to the plates for 2 h at room temperature. After washing, HRP-conjugated goat anti–mouse IgG1 antibody (Southern Biotech) was applied to the plates for 2 h at room temperature. Bound secondary antibodies were detected by incubation with SureBlue TMB 1–Component Microwell Peroxidase Substrate (KPL). The concentration of anti-NP antibodies was calculated by standard curve.

Means of two groups were compared with nonpaired two-tailed Student’s t test, and means between three or more groups were compared with one-way ANOVA with Bonferroni’s post-test using Prism software (GraphPad Software).

All microarray data are available in the Gene Expression Omnibus (GEO) database under accession no. GSE56883.

Fig. S1 shows generation of the S1pr2^Venus mice, and representative flow cytometry plots and gating strategy for the data shown in Fig. 1 (B–D). Fig. S2 shows representative flow cytometry plots and gating strategy for the data shown in Fig. 2 C. Fig. S3 shows gating strategy and representative flow cytometry plots for the data shown in Fig. 4. Fig. S4 shows representative histological images and supporting data for Fig. 5. Table S1 shows Entrez gene IDs and probe set IDs of genes described in Fig. 4 A. Videos 1 and 2 show dynamics of OT-II T cells of the indicated genotypes in the GC and FM. Video 3 shows tracking of OT-II T cells that entered the GC-FM interface zone.

We thank K.M. Ansel, M. Kitano, K. Kometani, S. Fagarasan, T.G. Phan, and T. Kurosaki for discussion, S. Kawamoto for anti-GFP staining, and IMS-RCAI core facilities for microarray analysis and cell sorting.

This work was supported by Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists (S. Moriyama), RIKEN Special Postdoctoral Researcher Program (S. Moriyama), and the Ministry of Education, Culture, Sports, Science, and Technology of Japan (T. Okada).

The authors declare no competing financial interests.