|Home | About | Journals | Submit | Contact Us | Français|
OX40 is a member of the Tumor Necrosis Factor Receptor Family expressed on activated and regulatory T (Treg) cells. Using an Ox40-cre allele for lineage marking, we found that a subpopulation of naïve T cells had also previously expressed OX40 in the thymus. Ox40-cre was induced in a small fraction of thymocytes that were OX40+, some of which were CD25hi Treg precursors. Thymic OX40 expression distinguished cells experiencing a strong signaling response to positive selection. Naïve T cells that had previously expressed OX40 demonstrated a partially activated phenotype that was distinct from that of the majority of naïve T cells. The results are consistent with the selection of Treg cells and a minor subpopulation of naïve T cells being dependent on strong signaling responses to thymic self ligands.
Signaling through the T cell antigen receptor (TCR) determines the fate of immature T cells at the CD4+CD8+ double-positive stage of their development in the thymus (1–3). Cells die at this stage when the magnitude of the TCR signaling they experience falls outside the range that is permissive for positive selection (4–7). Strong signals generated by TCRs that bind with high affinity to peptide/MHC ligands induce apoptosis by negative selection, whereas weak signaling, or an absence of it, also induces apoptosis. Thymic selection of this sort enriches the naïve T cell repertoire with TCR specificities that are likely to be useful in immune responses, and therefore its regulation is essential for adaptive immunity.
Commitment of cells to the CD4, CD8, NKT or regulatory T (Treg) cell lineages occurs coordinately with positive selection as cells move beyond the double-positive stage of development (8, 9). Signaling through the TCR allows for this commitment to take place, but whether it might be instructive for it has been the subject of intense investigation and controversy. In particular, strong versus weak TCR signals have been linked to commitment to the CD4 versus CD8 lineages respectively (10). This follows from the fact that p56Lck is more avidly associated with CD4 than it is with CD8 (11). CD4/TCR signaling in response to binding of MHC class II ligands might therefore be expected to be stronger than CD8/TCR signaling in response to binding of MHC class I ligands. That lineage choice is a function of this difference has, however, been challenged by conflicting findings that prompt an alternative model (12, 13).
The possible involvement of TCR signal strength in specifying the Treg lineage followed initially from observations that the representation of Treg cells relative to conventional T cells is increased in mice that carry transgene-encoded TCRs specific for self-antigens (14–16). Detailed analysis of the Treg cell repertoire subsequently suggested that it is distinct, albeit at least partially overlapping, with that of conventional CD4+ T cells (17). Consistent with this, the TCRs expressed by Treg cells were found to be enriched for autoreactivity (17). Independent TCR repertoire analysis and different assays for autoreactivity have prompted alternative interpretations concerning the distinctiveness and reactivity of the Treg cell repertoire (18). An additional recent complication comes from the observation that commitment to the Treg cell lineage can be influenced by conditioning of CD4−CD8− double-negative thymocytes by CD4+CD8+ double-positive cells (19) in what may be a lymphotoxin-dependent fashion (20). Such conditioning would obviously precede TCR signaling during positive selection and thus could call into question the primacy of the TCR signal in determining the Treg cell fate.
OX40 is a member of the Tumor Necrosis Factor Receptor family of proteins that is induced on activated T cells, and is constitutively expressed on peripheral Treg cells (21, 22). To allow for marking and mutagenesis of these cells, we generated a mouse carrying an insertion of the gene for the Cre recombinase in the Ox40 locus. Cre recombination in these mice occurs in the expected cell types with the pointed exception of a subpopulation of naïve T cells. Here we show that these cells are derived from a minor fraction of TCRhi thymocytes that expresses OX40. Like Treg cell precursors, the development of these cells is associated with strong signaling responses during their selection. We show that Ox40-cre allows for marking of cells as a consequence of distinctive signaling experiences in the thymus and that the marked cells exhibit lasting differences from the majority of naïve T cells. Lineage marking with Cre alleles such as Ox40-cre is a powerful means of resolving differences in populations of lymphocytes and linking them to distinct patterns of gene expression during crucial developmental periods.
A modified form of the cloning vector pSP72 was generated by ligating an adapter between the Xho I and Bgl II sites in the polylinker. The oligos used to create the adapter were: 5′-tcgagggatccgtcgacgcggccgcgaattcgagctca-3′ and 5′-gatctgagctcgaattcgcggccgcgtcgacggatccc-3′. The modified polylinker in the new plasmid (pSP40) contained the following arrangement of restriction enzyme sites: Xho I-BamH I-Sal I-Not I-EcoR I-Sac I-Bgl II. A 1.7Kb Hind III fragment of the Ox40 locus containing exons 1, 2 and most of 3 was blunted (using the Klenow fragment of E. coli DNA polymerase I) and ligated between the blunted Xho I and BamH I sites of pSP40 such that the upstream Hind III site was recreated by joining to the polylinker Xho I site. A Sal I-Not I fragment containing the mengovirus IRES (23) was then ligated between the Sal I and Not I sites of the plasmid (i.e., downstream of the genomic DNA) to create pSP40B. In separate ligations, an Mlu I fragment containing the NLS-Cre open reading frame from pMC1-cre (24) was inserted in an Spe I site upstream of the trimerized SV40 polyadenylation signal in a pBluescript derivative of the vector pSVA3 (using the Klenow fragment to blunt both vector and insert DNA). A Not I-NLS-cre-pA-EcoR I fragment from this plasmid was then inserted between the Not I and EcoR I sites of pSP40B to create pSP40C. This plasmid was further modified by insertion of an EcoR I-FRT-PGK-neo-pA-FRT-Sac I fragment (from a plasmid kindly provided by Dr. Kevin Jones, University of Colorado, Boulder) between the EcoR I and Sac I sites downstream of the Cre-pA DNA. The resultant plasmid (pSP40D) contained a Hind III -flanked insert comprised of 1.7Kb of Ox40 genomic DNA fused to an IRES-Cre-FRT-PGK-neo-pA-FRT element. This Hind III fragment was excised and inserted between Kpn I-Hind III and Hind III-Sac I fragments from the Ox40 locus that had been previously joined together between the Kpn I and Sac I sites of pBluescript-KS. The structure of the final construct was confirmed by restriction enzyme digestion prior to linearization and transfection into 129/Sv ES cells.
ES cells were screened by Southern blot using a BamH I-Sac I fragment of genomic DNA from the 3′ end of the locus. Clones that contained the mutation were expanded and microinjected into C57BL/6 blastocysts. Chimeric males were crossed to C57BL/6 females to achieve germline transmission of the mutation. A carrier mouse was subsequently crossed to a mouse carrying a Beta-actin-Flp transgene (25) to remove the FRT-flanked neo gene. neo-negative, cre-positive mice were then backcrossed to C57BL/6 partners for greater than 10 generations. At various points in the backcross, the Ox40-cre mice were crossed to mice carrying the Rosa26-loxP-STOP-loxP-YFP allele (26). All mice were maintained and bred under specific pathogen-free conditions under the approval of the UCSF Institutional Animal Care and Use Committee.
Conjugated antibodies were purchased from BD Biosciences, Invitrogen, and eBioscience. Phosphotyrosine and phospho-S6 ribosomal protein were detected with fluorochrome-conjugated P-Tyr-100 and D57.2.2E respectively from Cell Signaling Technologies. Ox40 was detected in three steps with biotinylated anti-OX40 (OX-86) followed by biotinylated anti-rat IgG1 (RG11/39.4) and then streptavidin-fluorochrome conjugates.
Single-cell suspensions were prepared from tissues using 0.45μm cell strainers (Falcon) and PBS containing BSA (0.3%, w/v). Cells were washed by centrifugation and then incubated on ice for 20–30 minutes with saturating concentrations of antibodies specific for cell surface molecules. After washing in PBS/BSA, the cells were either analyzed directly by flow cytometry, or were processed further for detection of intracellular molecules. DAPI (0.3μM) was used for live-dead cell discrimination whenever feasible. LACK/I-Ad tetramer staining was performed as previously described (27).
For FoxP3 or cytokine detection, cells were permeabilized with Foxp3 Fixation/Permeabilization Concentrate and Diluent (eBioscience) or Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (BD Biosciences) respectively. Cytokines were detected after a four-hour incubation in PMA (20ng/ml; Sigma) ionomycin (500ng/ml; Sigma) and Brefeldin A (1μg/ml; Epicentre Biotechnologies). For detection of phosphorylated proteins, single cell suspensions were prepared on ice and fixed immediately in PBS containing 2% paraformaldehyde for 15 minutes at 4°C. Control samples were treated with 1mM pervanadate or 20nM rapamycin at 37°C for 5 minutes prior to fixation. The cells were washed with PBS, incubated for 25 minutes on ice with 90% ice-cold methanol, then washed at least three times before incubation with antibodies directed against cell surface and intracellular antigens. All flow cytometry was performed using FACSCalibur and LSR II instruments (Becton Dickinson). Cell Sorting was performed using a FACSAria (Becton Dickinson) to a final cell purity of ≥95%.
Frozen pellets of sorted cells were incubated in 10mM Tris (pH8.0) containing 1mg/ml Proteinase K (Roche) for 1 hour at 55°C, then 10 minutes at 95°C. Approximately 50,000 cell equivalents were used for each real-time PCR. TREC primers: 5′-TTGCCTTTGAACCAAGCTG-3′; 5′-GAGCATGGCAAGCAGCAC-3′. FAM-labeled TREC probe: 5′-CACCTGCACCCTATGCATAAACCCACA-3′. Control primers: 5′-TTTGTCAAACACATGAGCACCT-3′; 5′-TCATTCTACGGGCAGTGTTG-3′. FAM-labeled control probe: 5′-AGACACTTACACGATCGATCCAAATGTGACA-3′.
CD4+CD25−CD44loYFP− and CD4+CD25−CD44loYFP+ cells were sorted from pooled spleen and lymph node cells from 8 donor mice. 450,000 sorted cells were injected into each Rag1−/− or C57BL/6 recipient mouse.
Thymocytes or sorted YFP− pooled spleen and lymph node cells from W15αβ TCR transgenic Ox40-cre/YFP mice were incubated in 48-well plates at 2 or 3×106/well respectively in the presence of LACK peptides. 2×106 Thy-1-depleted spleen cells were included in the thymocyte cultures. Cells were analyzed for YFP and activation marker expression at 24 and 72 hours. The LACK (156–173) peptide (28) was as follows: ICFSPSLEHPIVVSGSWD. Variant peptides had the following substitutions: H164N; H164Q; V167A; V169A.
Sorted cells (2 × 104 per well, U-bottom 96-well plate) were incubated with γ-irradiated (2000 rad) spleen cells (8 × 104 per well) and anti-CD3 (clone 145 2C11, 2μg/ml) for 72 hours. 3H-thymidine (0.5 μCi/well; Du Pont/NEN) was added during the last 4 hours of culture. Background counts in wells containing APCs alone were always <1,000 cpm.
RNA was prepared from approximately 1 million flow-sorted CD4+CD25+YFP+, CD4+CD25−CD44loYFP−, CD4+CD25−CD44loYFP+, CD4+CD25−CD44hiYFP+ cells using the RNeasy mini kit (Qiagen) with on-column DNAse digestion. Sample preparation, labeling, and array hybridizations were performed according to standard protocols from the UCSF Shared Microarray Core Facilities (http://www.arrays.ucsf.edu) and Agilent Technologies. Total RNA quality was assessed using a Pico Chip on an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). RNA was amplified and labeled with Cy3-CTP using the Agilent low RNA input fluorescent linear amplification kits following the manufacturer’s protocol. Labeled cRNA was quantified by spectrophotometry then hybridized to Agilent whole mouse genome 4x44K Ink-jet arrays. Arrays were scanned using the Agilent microarray scanner and raw signal intensities were extracted with Agilent Feature Extraction v9.1 software. The data were subsequently normalized using the quantile normalization method (29). No background subtraction was performed. A linear model was fit to each comparison to estimate means, and calculate moderated t-statistics, B statistics (30), false discovery rates (31) and p-values (32) for each microarray feature. All procedures were carried out using functions in the R package limma in Bioconductor (http://www.bioconductor.org (33)). All four populations mentioned above were purified from cells taken from single mice. Two mice were used to give eight samples (two of each population) and each sample was hybridized to a separate microarray. Gene ontology (GO) analysis was performed using DAVID (the Database for Annotation, Visualization, and Integrated Discovery (34)). The microarray data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus (35) and are accessible through GEO Series accession number GSE13502 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13502).
To express the Cre recombinase in activated T cells, we inserted an IRES-cre element into the third exon of the Ox40 gene in embryonic stem cells (supplemental Fig. S1). The linked FRT-flanked neomycin resistance gene in the targeted locus was removed from the germline of mice carrying this insertion by crossing them to a β-actin-Flp mouse (25). We then backcrossed the Ox40-cre mice to C57BL/6 mice before crossing them to other mice carrying a Cre-dependent YFP reporter allele (Rosa26-loxP-STOP-loxP-YFP (26)) to create Ox40-cre/YFP mice for use in experiments. Because insertion of the Cre open reading frame into the Ox40 locus destroys its capacity to produce OX40-encoding transcripts, we routinely analyzed mice that were heterozygous for the Ox40-cre allele and could therefore express haploid levels of OX40 from the wild-type allele they carried. Loss of one copy of the Ox40 gene does not detectably impair immune responses or the representation of lymphocyte subpopulations, so heterozygous Ox40-cre mice were expected to behave immunologically as if they were wild-type (36).
OX40 mRNA is expressed in the testes (37). We found that this was a consequence of expression in germ cells because male mice carrying both the Ox40-cre and YFP alleles passed the latter to their offspring in the recombined state. Similar results were obtained with other loxP-containing alleles (data not shown). Germline recombination could be avoided simply by ensuring that male parents did not contain Ox40-cre and a target (loxP-containing) allele. Neither OX40 nor Ox40-cre is detectably expressed in the female germline (37).
CD4+ YFP+ cells in the periphery (Fig. 1), be it in the lymph nodes, spleen or tissues could be classified as regulatory (FoxP3+), memory (CD44hi) or naïve (CD44lo). A very small fraction of CD8+ T cells were also YFP+ as might be expected given weaker induction of OX40 in activated CD8+ compared to CD4+ T cells (37). YFP+ CD8+ T cells were, however, present in much increased numbers in mice that had been infected with viruses (Klinger et al., in preparation). The penetrance of recombination in uninfected mice was reproducibly ~90% in Treg cells, 60–70% in CD44hi CD4+ T cells, 8–15% in CD44lo CD4+ T cells, and 2–5% in CD44hi CD8+ T cells. With the exception of CD44lo CD4+ T cells, this hierarchy correlated with the degree to which OX40 is induced in the different populations in the steady-state, or following activation.
Expression of YFP in CD44loCD25− peripheral T cells was initially unexpected because OX40 expression is induced on naïve T cells following their activation by antigen, and there is normally no significant expression of the molecule on cells that have a naïve phenotype (Fig. 1). We considered three possible circumstances in which Ox40-cre might be induced and cause YFP expression on CD44lo cells. One was that the peripheral YFP+ CD44lo cells might be derived from cells that acquired YFP in the thymus. Another was that the CD44lo YFP+ cells were in fact derived from CD44hiYFP+ cells that had simply reverted to a CD44lo phenotype. Finally, there was also the possibility that Ox40-cre expression was not tightly regulated in naïve T cells and that some of the cells stochastically passed through an OX40+ state in an inappropriate fashion that did not reflect the normal pattern of activation-induced OX40 expression.
To address the possibility that the CD44lo YFP+ cells acquired YFP expression in the thymus, we tested them for the presence of TCR rearrangement excision circles (TRECs (38)) using a real-time PCR assay. These circles are generated during rearrangement of the Tcra locus at the double-positive stage of development. Although they are stably maintained in naïve T cells, TRECs are not replicated during mitosis and consequently they are extremely infrequent in memory cells that have undergone extensive proliferation (39). CD44loYFP+ cells showed a similar TREC burden to CD44loYFP− naïve T cells (Fig. 2A) indicating that the YFP marker was acquired by the cells in the absence of mitosis. These data are not readily compatible with the YFP+CD44lo cells being derived from CD44hi cells, and they suggest instead that the YFP mark was acquired in the thymus during positive selection, or in the periphery in the absence of proliferation. Consistent with this, we also found that like naïve T cells – and in contrast to CD44hi T cells – the CD44loYFP+ cells failed to produce cytokines when they were activated briefly in vitro (Fig. 2B).
YFP-expressing cells were present in the thymuses of neonatal mice at frequencies that were equivalent to those of adult mice (data not shown). We could also detect YFP+ CD4+ T cells in the spleens of very young mice (Figures 2C and D) where the majority of them had a CD25−CD62LhiCD44lo phenotype. Very few YFP+ cells had a regulatory or memory phenotype at this age, although this had begun to change by 11 days after birth. Strikingly, the percentage of YFP+ cells in the CD44lo population did not change substantially with age (Fig. 2E) consistent with them being generated throughout life and representing a stable component of the naïve repertoire. By contrast, the proportions of the regulatory and memory populations that were YFP+ increased with age (Fig. 2E).
Ox40-cre could be induced in purified naïve YFP− CD4+ T cells as a consequence of proliferation following transfer into Rag1−/− recipients (supplemental Fig. S2A). This result fit with published data showing OX40 expression on T cells in such circumstances (40). Importantly, very few naïve T cells became YFP+ following transfer into lymphocyte-replete syngeneic recipients (supplemental Fig. S2B) and the few that did become YFP+ were clearly also high for expression of CD44. By contrast, the majority of CD44loYFP+ T cells retained their CD44lo phenotype following transfer into lymphocyte-replete recipients (supplemental Fig. S2A). Thus, although Ox40-cre can be induced in peripheral T cells following adoptive transfer, such induction is invariably associated with upregulation of CD44. From this we conclude that it is unlikely that the CD44lo YFP+ cells acquired their YFP mark in the periphery, and that they are instead cells that underwent Ox40-cre-mediated recombination during positive selection in the thymus.
Consistent with the normal pattern of OX40 expression in the thymus, YFP could be detected on less than 1% of thymocytes in Ox40-cre/YFP mice (Fig. 3A). The majority of these were single-positive CD4+CD8− cells but there were also small numbers of CD4+CD8+ and CD4−CD8+ cells (Fig. 3B).
Using OX40-deficient mice as staining controls, we could identify two subpopulations of OX40-expressing TCRhi cells (Fig. 3C): one that expressed intermediate levels of OX40 and no, or very little, CD25; and one that expressed higher levels of OX40 that was also CD25hi. Both of these subpopulations contained cells that had undergone Ox40-cre-mediated recombination (Fig. 3D) although the frequency of YFP+ cells was 5–6 fold higher in the latter than in the former consistent with the level of OX40 expression they displayed.
CD25hi thymocytes include FoxP3+ cells and their precursors (41). To determine whether OX40 upregulation might precede that of CD25, we analyzed thymocytes that expressed intermediate levels of YFP (Fig. 3E). Transcriptional output from the Rosa26 locus is normally uniform in all thymocytes, so intermediate levels of YFP discriminate cells that have recently undergone Ox40-cre-mediated recombination but have not yet acquired peak levels of the reporter (42). Approximately one-third of the YFPint cells were CD25hi suggesting that in these cells OX40 upregulation followed that of CD25. We noted, however, that there were significantly more (approximately twice as many) CD25+ cells in the YFPhi population than in the YFPint population (Fig. 3E). The simplest interpretation of this observation is that some of the YFPhi cells acquired CD25 expression after they underwent Ox40-cre-mediated recombination (i.e., in these cells, OX40 expression preceded that of CD25). Alternative interpretations include proliferation of CD25hi but not CD25lo cells, different rates of apoptosis or exit for the two types of cells, or the fact that the number of CD25hi cells might be increased by inclusion of recirculating peripheral regulatory T cells (43). Collectively, these data indicate that OX40 expression is unlikely to provide additional resolution to CD25 for discriminating all Treg precursors.
In peripheral T cells, OX40 is induced as a consequence of TCR signaling (44). To examine the relationship between TCR signaling and Ox40-cre activity, we generated Ox40-cre/YFP mice that expressed a transgene-derived αβ TCR (the W15αβ TCR (27)) specific for an I-Ad-restricted peptide from the Leishmania major LACK protein. CD4+ T cells and thymocytes from these mice were stimulated in vitro with the cognate LACK peptide and with variants of it (containing alternative residues at presumptive TCR contact positions (45)). As shown in Fig. 4A, the frequency of YFP+ cells in the cultures of both kinds of cells increased as a function of antigen concentration. Moreover, we found that strongly agonistic ligands of the LACK/I-Ad-specific TCR (Fig. 4B) induced considerably more recombination than did similar concentrations of weakly agonistic ligands (Fig. 4A).
If Ox40-cre is induced in thymocytes as a function of TCR signal strength, then OX40+ cells in the thymus should be distinguished by stronger responses to selecting ligands than the majority of thymocytes that fail to induce OX40. TCR signaling in double-positive thymocytes results in the phosphorylation of tyrosine residues on multiple proteins (46). TCRhi double-positive cells that are actively undergoing positive selection should therefore be distinguishable from pre-selection (TCRlo) cells on the basis of their content of intracellular phosphotyrosine. As shown in Fig. 4C, this difference could be revealed by flow cytometry with a monoclonal antibody specific for phosphotyrosine. TCRhi cells expressing OX40 were enriched for high phosphotyrosine content relative to OX40−TCRhi cells, and this was true regardless of whether they also expressed CD25 (Figs. 4C and 4D). Thus, OX40 (and/or CD25) expression discriminates a population of thymocytes experiencing a strong TCR-proximal signaling response to selection ligands.
Phosphorylation of the 32kDa S6 ribosomal protein is mediated by AGC kinases that lie downstream of PDK1 activation (47, 48), which is itself subject to regulation by TCR/CD28 signaling (49, 50). S6 phosphorylation is decreased in thymocytes that lack PDK1, and such thymocytes are severely compromised in their development (51). We noted that like phosphotyrosine, positive selection and the upregulation of TCR levels could be correlated with higher levels of intracellular phospho-S6 (Fig. 4C). Furthermore, OX40 expression (with or without CD25) discriminated a subpopulation of TCRhi cells with the highest content of phospho-S6 (Fig. 4D).
Additional evidence of OX40 expression correlating with a strong signaling response came from the analysis of cell surface molecules that change in expression in response to peptide/MHC-dependent selection (52). OX40+ CD4+ thymocytes showed high expression of PD-1, CD44, CD53 and CD5 (Fig. 4D). Interestingly, the CD25− subpopulation was marked by higher expression of CD69 than the CD25+ one, probably reflecting the presence of more mature or recirculating (43) cells in the latter population than the former.
Collectively, the results just summarized show that OX40 and CD25 expression discriminate two subpopulations of thymocytes (Fig. 3C), both of which are distinguished by the strength of their responses to selection ligands. Although many of these cells are the precursors of peripheral Treg cells, others are the presumptive precursors of the CD44lo cells that express YFP in Ox40-cre/YFP mice.
Prior Ox40-cre induction was extremely infrequent in naïve T cells expressing a transgene-encoded αβ TCR on their surfaces (supplemental Figs. S3A and S3B; W15αβ mice (27), CD44lo LACK/I-Ad+ cells). CD44lo cells in the same mice that had displaced the transgenic TCR from their surfaces (due to rearrangement of endogenous TCRα chains) included normal numbers of YFP+ cells (supplemental Fig. S3B; W15αβ mice, CD44lo LACK/I-Ad− cells). These observations suggested that Ox40-cre induction during positive selection depends on the specificity of the TCRs expressed by thymocytes, and that the αβ TCR we used did not allow for appropriate signaling during positive selection for this to occur.
Similar observations were made using mice with a diverse repertoire that was constrained by expression of a transgene-encoded TCRβ chain (supplemental Figs. S3A and S3B; W15β mice (27)). LACK/I-Ad-reactive naïve T cells were present in increased numbers in these mice relative to nontransgenic controls (because of the specificity of the fixed TCRβ chain). These cells, included YFP+ cells, but at much reduced frequencies relative to cells that were not LACK/I-Ad-reactive (supplemental Fig. S3B).
Finally, we noted that YFP+ and YFP− populations of CD44lo T cells differed from one another in their expression of individual TCR variable domains (supplemental Fig. S3C). Collectively, these observations indicated that thymic OX40/Ox40-cre induction identifies a population of naïve T cells with a different TCR repertoire from other naïve T cells. This conclusion would be predicted by the observations that thymic OX40 expression discriminates cells undergoing distinct signaling experiences during selection.
As an additional means to assess the relationship between YFP+CD44lo cells and other peripheral CD4+ T cell populations (YFP− naïve cells, CD44hi memory cells, and YFP+ CD25+ regulatory cells), we examined the genes they express using microarrays. Two-way comparisons between the four populations revealed multiple genes that were differentially expressed by ≥1.5 fold (Fig. 5A and supplemental Table S1). The largest numbers of such genes were found in comparisons between naïve T cells and either memory or regulatory cells (2958 and 2283 genes respectively). By contrast, only 782 genes were differentially expressed between naïve and CD44loYFP+ cells. Strikingly, however, there were 963 and 1283 genes that were differentially expressed between the CD44loYFP+ population and either regulatory or memory cells respectively. This superficial comparison suggested that the CD44lo YFP+ population was most similar to naïve T cells, but that it also shared aspects of its gene expression pattern in common with both regulatory and memory cells.
From the entire microarray dataset, we identified 4109 genes that were differentially expressed in at least one of all possible two-way comparisons between the four populations. This set of genes is represented in Fig. 5B ordered according to relative expression in the CD44hi and naïve populations. As a further means for assessing relatedness between the four populations, we calculated Pearson correlation coefficients for the six possible pairwise comparisons between them with respect to their expression of the 4109 genes. The highest correlations (lowest distances as graphed in Fig. 5C) were evident in comparisons between CD44loYFP+ (L) and either naïve (N) or regulatory (R) T cells. By contrast, naïve T cells compared with either regulatory or memory (H) T cells showed substantially lower correlation coefficients.
Naïve T cells are quiescent cells marked by small amounts of cytoplasm and low transcriptional activity compared to memory T cells (53). Genes that are upregulated in the latter (supplemental Table S2; 2254 (H/N)hi genes) include those that are associated with lymphocyte activation and proliferation as opposed to a small subset of genes involved in specific metabolic processes that are more highly expressed in naïve T cells (supplemental Table S2; 1855 (H/N)lo genes). We recalculated Pearson correlation coefficients for the six comparisons focusing independently on these up- or down-regulated genes. The three comparisons involving CD44loYFP+ T cells (L-N, L-R, and L-H) gave similar correlation coefficients to one another when the 2254 (H/N)hi genes were used (Fig. 5C, right). By contrast, when the 1855 (H/N)lo genes were used for the calculations, the L-N comparison showed the highest correlation coefficient (Fig. 5C, middle). This analysis therefore suggested that the CD44loYFP+ population shared aspects of its gene expression profile in common with all three types of T cells to which it was compared, but it was especially similar to naïve T cells in its expression of genes whose high expression distinguishes the naïve state.
Genes differentially expressed between CD44loYFP+ cells and naïve T cells were enriched for annotation terms associated with T cell activation (supplemental Table S3). By contrast, annotation terms associated with proliferation, apoptosis and signaling were more enriched among the genes that were differentially expressed between regulatory and memory T cells compared with naïve T cells. This type of analysis suggested that the CD44loYFP+ cells should show indications of activation. We confirmed this by flow cytometry focusing in particular on activation markers that were identified as being differentially expressed between the YFP+ and YFP− subpopulations of naïve T cells by microarray analysis. Representative data from these experiments are shown in Fig. 6. A key observation was that CD44lo YFP+ cells were distinct from the majority of naïve T cells (i.e., the CD44lo YFP− cells) most notably in their expression of Ly-6C (detected with the AL-21 mAb). 20–30% of the cells showed elevated expression of CD69 and reduced expression of CD62L. CD83, CXCR3, and CD5 levels were also reproducibly elevated on the cells (Fig. 6 and data not shown).
Less than one percent of thymocytes express OX40, and these cells are primarily CD4+CD8− in phenotype. About a quarter of them express CD25 and include FoxP3+ Treg cells and their precursors. Among the remainder, some are the presumptive precursors of a subpopulation of naïve T cells that is distinguished by its gene expression profile, cell surface phenotype, and by its TCR repertoire. The most distinctive characteristic of OX40+ thymocytes is the fact that they are enriched for indications of a strong signaling response to thymic selection ligands. Thus, OX40 is a useful marker for cells that fall at the top end of the distribution of responses to thymic self antigens, and consequently OX40 induction could represent a useful means for identifying cells that are of significance either for suppressing or potentiating autoimmunity.
The Ox40-cre allele we have used in this study is induced in response to TCR stimulation in a similar fashion to that of OX40. It is more active in CD4+ than CD8+ T cells, and its level of induction is a function of the strength of the TCR stimulus a cell experiences (although it is likely that other signals can also influence its expression (54)). We show that strongly agonistic ligands of a LACK/I-Ad-specific TCR induced more Cre recombination than ligands that were weakly agonistic; we have also made similar observations with a panel of ligands for a class I-restricted TCR expressed on CD8+ T cells (data not shown). The relationship between strength of stimulus and amount of Ox40-cre induction predicted that the few thymocytes expressing OX40 would be enriched for those making strong responses to selecting ligands. Consistent with this, phosphotyrosine and phospho-S6 levels were both elevated in OX40+ cells, and they showed a cell surface phenotype expected of cells that are strongly stimulated.
The analysis of mice that feature constrained TCR repertoires has revealed differences but substantial overlap between regulatory and conventional TCR repertoires (17, 18, 55, 56). The overlap is inconsistent with commitment to the Treg lineage being solely a consequence of strong reactivity to thymic self ligands. It is, however, permissive of models of Treg cell development in which strong TCR signaling potentiates the Treg fate in a probabilistic fashion and/or in combination with other factors (41, 57). Conversely, it allows for the formation of naïve T cells from thymocytes that might have experienced the same magnitude of TCR signaling response as Treg precursors. OX40 is a potential marker for the latter class of thymocytes based on its induction as a function of TCR signal strength.
Regulatory T cells are a prominent product of the thymus in mice that carry transgenes encoding TCRs with high affinity for self ligands (14–16). This finding is suggestive of a potential instructive role for strong recognition of self in commitment to the regulatory lineage. In support of this, Hsieh et al. found that TCRs cloned from Treg cells were associated with enhanced capacity to drive T cell expansion in lymphopenic hosts, or to induce autoreactive proliferative responses in vitro (17). Although these results have been challenged by other work that failed to show evidence of self-reactivity in the Treg TCR repertoire (18), they provide a straightforward explanation for the observed differences in the Treg and conventional TCR repertoires (i.e., that the former contains more strongly self-reactive specificities than the latter). Moreover, they are consistent with observations presented here that thymocytes marked by OX40 and/or CD25 upregulation are enriched for cells making stronger signaling responses than cells that lack expression of these molecules.
Intrathymic injection experiments have shown that FoxP3−CD25hi thymocytes include precursors of Treg cells (41). Upregulation of FoxP3 in these cells can occur in a fashion that is independent of persistent engagement of TCR ligands, and it correlates with a capacity to signal productively in response to IL-2 (57). Notably, not all CD25hi cells induce FoxP3 subsequent to IL-2 treatment, and similarly, a substantial fraction of them do not become FoxP3+ after intrathymic transfer (41). The majority of CD25hi thymocytes coexpresses OX40, and therefore should be marked by Ox40-cre. Hence, at least some of the YFP+ CD44lo T cells in Ox40-cre/YFP mice are derived from CD25hi thymocytes, these likely being the cells that do not become Treg cells following intrathymic transfer (41).
Sub-maximal expression of YFP from the ROSA26-YFP allele identifies thymocytes that have recently undergone Cre-mediated recombination (42). Our analysis of this population did not reveal clear evidence that OX40 upregulation depends on prior CD25 upregulation or vice versa. We noted that the frequency of CD25hi cells was higher in the YFPhi population than in the YFPint population, an observation that could be consistent with upregulation of CD25 occurring after that of OX40 in some cells. As mentioned, however, there are other possible explanations for this. Intrathymic injection experiments could help to resolve what fraction of OX40+CD25− thymocytes might become CD25+, but such experiments would not readily address the more compelling question of whether any of the YFP+CD44lo cells are derived from OX40+ cells that do not transition through a CD25hi state. This question is compelling simply because answering it would help to clarify the developmental paths that lead to the YFP+CD44lo population and the points at which they might diverge from the path to Treg cells.
The cell surface phenotype of YFP+ CD44lo T cells was different from that of YFP− CD44lo T cells and suggestive of ongoing differences in their responses to self-ligands, and/or a persistent imprint of their distinct thymic experiences. In particular, we noted enrichment for high expression of Ly-6C, which has previously been identified as a marker for naïve T cells that are proficient at providing help for plasma cell differentiation (58). Ly-6C has been implicated in lymphocyte homing (59) suggesting that the YFP+CD44lo T cells might exhibit different migration patterns in the body from other naïve T cells (although we have not yet detected such differences). Other changes, such as elevated CD69, and decreased CD62L on some of the YFP+ cells were suggestive of partial activation (partial because the cells were low for CD44 expression).
Microarray profiling provided a deeper perspective on differences between the YFP+ and YFP− CD44lo cells, and yielded a substantial collection of differentially expressed genes. This collection was enriched significantly for genes associated with lymphocyte activation consistent with the flow cytometry results just mentioned. Although some of the differentially expressed genes could be attributed to the presence of a small number (~5% of the total) of FoxP3+ cells within the YFP+ population, most of them had a distinct origin and were indicative of true differences between the two populations. In addition to the value of these experiments in identifying differences, they were also of use in assessing the relatedness of the YFP+ CD44lo cells to naïve (i.e., YFP− CD44lo), memory, and regulatory cells. Here again, we noted that the YFP+ cells shared aspects of their gene expression pattern in common with memory and regulatory cells but were nonetheless most closely related to naïve T cells. This was especially true when we focused our analysis on genes that are normally expressed at higher levels in naïve T cells compared to memory cells, and thus are likely to include genes essential for maintaining the naïve state.
In summary, the results presented here show that prior expression of a gene that is normally associated with peripheral T cell activation can be used to identify a distinct subpopulation of naïve T cells. These cells are expected to be enriched for expression of TCRs that have above-average reactivity for thymic self ligands because they induced strong signaling responses during positive selection. They are therefore of potential significance in the context of autoimmunity. The Ox40-cre allele is representative of a class of lineage marking alleles that in principle can be used to split the naïve T cell repertoire on the basis of differential thymic selection experiences. Such alleles should allow for correlations to be drawn between gene expression history and functional properties of T cells, and thus promise a unique perspective on the pathway to healthy and diseased immune responses.
A. Structure of the Ox40 locus showing translated (filled) and untranslated (open) exons and relevant restriction enzyme sites; the targeting vector and the mutant allele (after removal of the FRT-flanked neo gene are also shown. B. Southern blot results from the analysis of targeted ES cells (left) and tail DNA from mice (right) probed with the indicated 3′ probe and a full-length cDNA probe respectively. The neo gene was present in the ES cells but absent from all mice used in experiments.
A. The plots and histograms show donor-derived T cells in the blood of a Rag1−/− recipient (top) that received YFP−CD44lo CD4+ T cells, and a lymphocyte-replete (C57BL/6) recipient (bottom) that received YFP+CD44lo CD4+ T cells. B. The contour plots show YFP and CD44 expression on donor-derived T cells in the blood, liver and spleen following transfer of YFP−CD44lo CD4+ T cells into a lymphocyte-replete (C57BL/6) recipient. The histograms show overlays of CD44 expression on donor-derived T cells that are YFP+ or YFP−. Note that CD44hi YFP− T cells in these plots (and those in A.) are CD44hi cells that contaminated the sorted population, or CD44lo T cells that were activated following transfer but did not induce Ox40-cre recombination, or recipient T cells contaminating the analysis gate.
A. The contour plots show expression of CD44 and staining with a LACK/I-Ad multimer on cells from B10.D2 mice carrying a LACK/I-Ad-specific αβ TCR transgene (left) or a LACK/I-Ad-specific TCR β chain transgene (middle), or from a control mouse that was not transgenic for a TCR. B. Histograms show YFP expression on cells from the indicated quadrants of the contour plots. Spleen cells were used for the TCR transgenic mice, whereas blood cells were used for the B10.D2 control. C. Frequencies of CD44loYFP+CD4+ T cells expressing the indicated TCR variable domains in each of five mice (normalized to the mean frequencies of CD44loYFP− T cells expressing the same variable domains).
Four populations of cells (N, H, L and R) were sorted from two mice to give 8 samples. RNA prepared from these populations was analyzed using Agilent microarrays as described in Methods and Materials. The tables show genes that were differentially expressed (>1.5-fold difference in normalized expression value; B statistic>0) between the populations for each two-way comparison: L–N (A), L–R (B), H–L (C), R-N (D), H–N (E) and H–R (F). Columns in the tables show Agilent IDs, gene names, gene symbols, Genbank accession numbers, Unigene IDs, Entrez gene identifiers, mean normalized expression (log2 fluorescence) values for each population (H, L, N and R), the difference in expression values for the relevant comparison, and the associated B statistic.
Genes identified as differentially expressed (>1.5 fold difference, and B statistic >0) in two-way comparisons between L, H, N and R populations were ordered and divided into (H/N)hi and (H/N)lo sets of 2255 and 1854 genes respectively as shown in Fig. 5B. These genes were then analyzed for enriched gene ontology (GO) annotation terms (biological process category) relative to the entire mouse genome using the DAVID Functional Annotation tools (April, 2008 release). The table summarizes the terms that showed the highest degree of enrichment by modified Fisher Exact test ordered according to p-values (adjusted for multiple sampling). Count refers to the number of genes in the query list that contained the indicated GO term. Fold-change refers to the degree to which each GO term is over-represented in the query list relative to the mouse genome. The analysis of the (H/N)hi set of genes was restricted to terms at level 5 in the annotation tree.
The table summarizes gene ontology (GO) annotation terms that are enriched in at least one of the three indicated lists of differentially expressed genes (Fig. 5A) relative to the mouse genome. Fold change and p-values are as in Table S2. Shaded boxes signify a lack of enrichment for the indicated terms in a specific list.
We are indebted to Rebecca Barbeau of the UCSF Shared Microarray Core Facility for assistance with the microarray experiments. We thank Ming Hu, Jie Wie and Gavina Benitez for expert technical assistance, and members of the Killeen laboratory for helpful discussions. We thank Dr. Frank Costantini (Columbia University) for generously providing mice carrying the Rosa26-loxP-STOP-loxP-YFP allele.
1This work was supported by a grant from the National Institutes of Health (AI39506). NK was supported by a Scholar Award from the Leukemia and Lymphoma Society and MK was supported by a Natural Sciences and Engineering Research Council of Canada Postdoctoral Fellowship award.
The authors have no conflicting financial interests.