CTCF regulates cell cycle progression of αβ T cells in the thymus

Abstract

The 11-zinc finger protein CCCTC-binding factor (CTCF) is a highly conserved protein, involved in imprinting, long-range chromatin interactions and transcription. To investigate its function in vivo, we generated mice with a conditional Ctcf knockout allele. Consistent with a previous report, we find that ubiquitous ablation of the Ctcf gene results in early embryonic lethality. Tissue-specific inactivation of CTCF in thymocytes specifically hampers the differentiation of αβ T cells and causes accumulation of late double-negative and immature single-positive cells in the thymus of mice. These cells are normally large and actively cycling, and contain elevated amounts of CTCF. In Ctcf knockout animals, however, these cells are small and blocked in the cell cycle due to increased expression of the cyclin-CDK inhibitors p21 and p27. Taken together, our results show that CTCF is required in a dose-dependent manner and is involved in cell cycle progression of αβ T cells in the thymus. We propose that CTCF positively regulates cell growth in rapidly dividing thymocytes so that appropriate number of cells are generated before positive and negative selection in the thymus.

Introduction

The 11-zinc finger protein CCCTC-binding factor (CTCF) is a widely expressed and highly conserved transcriptional regulator implicated in many important processes in the nucleus (for reviews, see Ohlsson et al, 2001; Lewis and Murrell, 2004). In line with this view, murine CTCF is essential for early embryonic development (Fedoriw et al, 2004). CTCF is the archetypal vertebrate protein that binds insulator sequences, DNA elements that have the ability to protect a gene from outside influences (Bell et al, 1999). Its methylation-sensitive interaction with the imprinting control region of the H19/insulin-like growth factor 2 (Igf2) genes indeed controls enhancer access (Bell and Felsenfeld, 2000; Hark et al, 2000; Fedoriw et al, 2004). CTCF-mediated insulator activity has been predicted at several other sites including the DM1 locus and boundaries of domains that escape X-chromosome inactivation (Filippova et al, 2001, 2005). We have shown that CTCF mediates long-range chromatin interactions and regulates local histone modifications in the β-globin locus (Splinter et al, 2006). Evidence has furthermore been presented for a function of CTCF in inter-chromosomal interactions between Igf2 and other loci (Ling et al, 2006). During mitosis, CTCF remains bound to mitotic chromosomes, possibly facilitating reformation of higher-order chromatin loops after mitosis (Burke et al, 2005). Taken together, these data suggest that CTCF is an essential organizer of imprinting, long-range chromatin interactions and transcription.

Genome-wide mapping of CTCF-binding sites revealed ∼14 000 sites, whose distribution correlates with genes but not with transcriptional start sites (Kim et al, 2007). Strikingly, the 20-bp consensus motif found in the majority of the sites is virtually identical to a consensus sequence (called LM2^*), which is bound by CTCF and is found in ∼15 000 conserved non-coding elements (CNEs) in the human genome (Xie et al, 2007). Thus, CTCF-binding sites are part of CNEs, which are conserved across species and appear to have regulatory functions. High-resolution profiling of histone methylation in the human genome showed that CTCF sites mark boundaries of histone methylation domains (Barski et al, 2007) consistent with a function of CTCF as an insulator protein. Genome-wide analyses also revealed CTCF-binding sites near genes displaying extensive alternative promoter usage, including Protocadherin γ, the Immunoglobulin (IG) λ light chain and the T cell receptor (TCR) α/δ and β chain loci. In mice, CTCF-dependent insulators were found downstream of the Tcr α/δ and the Ig heavy chain loci (Magdinier et al, 2004; Garrett et al, 2005). Moreover, in T cells, CTCF-binding sites overlap significantly with DNAse I hypersensitive sites (HS), suggesting that CTCF is somehow involved in global T-cell gene expression (Boyle et al, 2008). These data imply an important function of CTCF in lymphocytes, in particular in the regulation of gene transcription and recombination targeting in complex loci.

T-cell progenitors differentiate in the thymus, where early precursors (lacking the cell surface markers CD4 and CD8 and therefore termed double-negative (DN) cells) develop into mature CD4 or CD8 single-positive (SP) T cells, following a regulated differentiation programme (see Supplementary Figure S1 for a schematic overview; for review, see Rothenberg and Taghon, 2005). The DN precursor population is generally subdivided into four distinct developmental stages (DN1–DN4), which are defined by differential expression of the cell surface markers CD25 and CD44 (interleukin-2 receptor and phagocyte glycoprotein 1, respectively). Further differentiation of T cells depends on rearrangement of the TCR gene segments of four loci, that is, α, β, γ and δ. Most mature T cells express an αβ TCR on their cell surface. These are the T cells involved in the classic adaptive immune response. γδ T cells represent only a small fraction of T cells in lymphoid organs of humans and mice (1–5%) and line the epithelial layer of various organs, including the small intestine, but their function remains somewhat enigmatic (for review, see Chien and Konigshofer, 2007). The choice of whether the α and β TCR genes are rearranged, or the γ and δ genes, is made at the DN2–DN3 stage in the thymus.

TCRβ gene rearrangement is initiated and completed at the DN3 stage. Upon productive (in-frame) rearrangement of the TCRβ gene, the TCRβ chain associates with the invariant pTα chain on the cell surface and forms the pre-TCR complex. Cells that have passed this so-called ‘β-selection' checkpoint are termed ‘β-selected' cells. The pre-TCR complex signals cells to proliferate and to downregulate CD25 expression. Cells subsequently acquire both CD4 and CD8 coreceptors to become double-positive (DP) cells, with CD8 usually being expressed first in most mouse strains (immature SP (ISP) cells). As a result, the late DN3, DN4 and ISP stages consist of large cycling cells (see Supplementary Figure S1). The DP cells then leave the cell cycle and rearrange their TCRα gene locus. If TCRα gene recombination is productive, TCRαβ is expressed on the cell surface of DP cells. TCRαβ-bearing immature cells are selected for major histocompatibility complex (MHC) recognition during the process of positive selection. TCRαβ receptors with specificity for MHC class I will develop into the CD8-positive (CD8⁺) T-cell lineage, whereas receptors recognizing MHC class II will become CD4-positive (CD4⁺) T cells. DP thymocytes that fail to recognize MHC class molecules die ‘by neglect', whereas potential self-reactive T lymphocytes are eliminated by a process called negative selection. When combined, these selection processes result in the generation of CD4 and CD8 SP thymocytes with TCRαβ receptors that can recognize non-self-antigens presented by MHC class II and I proteins, respectively. Mature SP cells exit the thymus and circulate to the periphery as naive CD4⁺ and CD8⁺ T cells.

To circumvent the problem of embryonic lethality (Fedoriw et al, 2004), we generated mice with a conditional Ctcf allele (Ctcf^f/f). We examined the potential function of CTCF in Tcr gene rearrangement and global T-cell gene expression by deleting Ctcf^f/f in thymocytes. Here, we show that CTCF exerts an effect as a critical regulator of cell growth and proliferation following β-selection in the thymus. We demonstrate that CTCF expression varies during normal T-cell differentiation, with the highest levels occurring in subpopulations of relatively large and cycling thymocytes, including ISP cells. Interestingly, knockout of Ctcf results in a cell cycle arrest at the ISP cell stage, owing to highly increased amounts of the cyclin-CDK inhibitors p21 and p27. CTCF-deleted DN4 and ISP cells are also significantly smaller than normal cells. We therefore propose a global function of CTCF as a positive regulator of cell growth in αβ T cells.

Results

Conditional deletion of the Ctcf gene in developing T lymphocytes

To understand CTCF function in vivo, we generated a conditional Ctcf allele (Ctcf^f) by inserting loxP sites upstream of exon 3 and downstream of exon 12 (Figure 1A). Equivalent levels of CTCF are expressed in Ctcf^f/f and wild-type mice (data not shown). Ctcf^+/f mice were crossed with mice expressing Cre recombinase ubiquitously (Sakai and Miyazaki, 1997). This causes removal of Ctcf exons 3–12 from Ctcf^f, yielding the Ctcf⁻ allele, in which a Ctcf–lacZ fusion transcript is expressed instead of Ctcf (Figure 1A). Ctcf^+/− mice appear normal and are fertile, but no homozygous knockouts are born from Ctcf^+/− crosses (Table I), consistent with an essential function of CTCF in early development (Fedoriw et al, 2004). Surprisingly, the ratio of wild-type to Ctcf^+/− littermates is higher than expected on a Mendelian basis in crosses among Ctcf^+/− mice and between wild-type and Ctcf^+/− mice (Table I). These data suggest that CTCF is required in a dose-dependent manner.

Figure 1.

Conditional targeting of the mouse Ctcf gene. (A) Murine Ctcf locus and gene targeting constructs. Exons of the Ctcf gene (solid boxes) are numbered, scale is in kilobase (K). Exon 1 is embedded in a CpG island. Exon 3 contains the start codon, and exon 12 contains the stop codon. Southern blot probes (5′-end and HIII) are shown above the Ctcf gene. The two targeting constructs, with loxP sites (small triangles), flanking a PMC1-neomycin cassette (neo^r) or a PGK-puromycin cassette (puro^r), are shown with homologous regions. TK, thymidine kinase gene; SA-LacZ, Splice acceptor-lacZ cassette (Hoogenraad et al, 2002). PCR primers for genotyping (p8563, p8946, p260 and p261, large triangles) are indicated on targeting cassettes. Underneath the targeting constructs, the deleted Ctcf gene is shown, which is generated after complete Cre-mediated recombination at the outermost loxP sites. Owing to alternative splicing, the splice acceptor (SA) site, present at the 5′-end of the reporter LacZ cassette, is spliced on to Ctcf exon 1 or 2, thereby generating a hybrid Ctcf–lacZ transcript. (B) Southern blot analysis of Lck–Cre recombinase activity. Digested genomic DNA from thymus and spleen of mice of the indicated genotypes was analysed by hybridization with the HIII probe (see panel A). The positions of the wild-type (WT), Ctcf^f/f (flox) and Ctcf^−/− (del) alleles are indicated. (C) Western blot analysis of thymus. Total thymus lysates from Lck–Cre Ctcf^f/f and WT mice (+ indicates presence of Cre transgene; − indicates absence) were analysed for CTCF, DNMT1 and UBF protein levels. Three mice were analysed per genotype. (D) Flow cytometric analysis of lacZ expression in CTCF conditionally deleted mice. LacZ expression was analysed in conjunction with cell surface markers. The indicated cell populations were gated and lacZ expression data are displayed as histogram overlays of Lck–Cre Ctcf^f/f mice (green) on top of background signals in wild-type mice (black).

Table 1.

Genotype of Ctcf^+/− × Ctcf^+/− and Ctcf^+/− × wild type offspring

Age	Genotype and number
	Wild type	Ctcf+/−	Ctcf−/−
Ctcf+/−× Ctcf+/−
E 9.5	13	14	0
E 3.5	10	7	0
Adult	88	92	0
Ctcf+/−× wild type
Adult	101	74	NA

To obtain a T-cell-specific deletion of the Ctcf gene, we crossed Ctcf^f/f mice with Lck–Cre transgenic mice, in which the proximal Lck promoter drives expression of the Cre recombinase (Lee et al, 2001; Wolfer et al, 2002). Southern blot analysis shows almost complete deletion of the Ctcf gene in thymus, whereas in spleen, deletion is not evident (Figure 1B). These data reflect the specificity of the Lck–Cre transgene; they also indicate that Ctcf knockout T cells do not populate the spleen in large numbers. To evaluate the onset of Ctcf^f gene deletion, we analysed lacZ expression in thymocytes by flow cytometry using fluorescein-di-β-D-galactopyranoside (FDG) as a substrate in conjunction with cell surface markers that define thymocyte subpopulations. We find that deletion is almost complete from the DN2 stage onwards (Figure 1D). Western blotting shows that in thymic nuclear extracts from Lck–Cre Ctcf^f/f mice, CTCF protein levels are reduced to ∼8% of control (Figure 1C). We conclude that ablation of the Ctcf gene results in an efficient depletion of the protein in vivo.

Defective TCRαβ lineage development in Lck–Cre Ctcf^f/f mice

To examine the effects of a Ctcf deletion, thymocyte subpopulations in 6- to 8-week-old mice were analysed by flow cytometry. This revealed that Lck–Cre Ctcf^f/f mice have a reduced thymic cellularity (Table II). Whereas a specific defect is observed in αβ T-cell development (Figure 2), γδ T cells are not affected by Ctcf deletion (Supplementary data and Supplementary Figure S2), similar to what has been reported for condtional deletion of DNMT1 (Lee et al, 2001) and the RNaseIII enzyme Dicer (Cobb et al, 2006). Apparently, cell division, chromatin organization and gene regulation in γδ T cells is quite different from αβ T cells.

Table 2.

Average numbers of thymocyte subpopulations derived from Lck–Cre Ctcf^f/f mice (homozygous), Lck–Cre Ctcf^+/f mice (heterozygous) and wild-type (WT) mice

Genotype	DN1	DN2	DN3	DN4	ISP	DP	CD4 SP	CD8 SP	Total
WT (n=14)	0.27±0.02	0.37±0.05	2.09±0.19	1.10±0.13	1.81±0.21	165±10.3	14.1±1.08	5.56±0.58	191±11.7
Heterozygous (n=7)	0.32±0.03	0.66±0.14	3.42±0.93	0.73±0.16	1.92±0.31	95.9±7.40	7.66±0.81	2.48±0.35	113±8.77
Homozygous (n=12)	0.39±0.08	0.91±0.26	4.72±0.85	3.15±0.40	9.14±1.82	35.4±1.89	2.19±0.50	0.81±0.20	56.7±3.76
P-value (heterozygous–WT)	NS	NS	NS	NS	NS	<0.0001	0.0001	0.0002	<0.0001
P-value (homozygous–WT)	NS	NS	0.0082	0.0002	0.0014	<0.0001	<0.0001	<0.0001	<0.0001
NS, not significant.
Average values are given in millions of cells±s.e.m.
P-values were calculated using a two-tailed student t-test.

Figure 2.

Defective TCRαβ lineage development in CTCF-deficient mice. (A) Representative flow cytometric analyses of cell populations from thymus (upper three panels) and spleen (lower panel) derived from wild-type (WT) and Lck–Cre Ctcf^f/f mice. Expression profiles of surface markers (indicated on the left) are shown as dot plots, and the percentages of cells within quadrants or gates are given. The DN1–DN4 cell populations (second panel from top) were gated on the basis of CD44 and/or CD25 expression, using the CD3⁻CD4⁻CD8⁻ triple-negative fraction as a basis. The third panel from the top shows gating of the ISP and CD8 SP cell populations (on the basis of low and high expression of CD3, respectively) using the fraction of CD4⁻CD8⁺ thymocytes (indicated in the CD4/CD8 profile of the upper panel). Note the accumulation of ISP cells in the thymus of the Lck–Cre Ctcf^f/f mouse (95%) compared with wild type (20%). (B) Number of thymic and splenic T cell subpopulations. Each symbol represents one individual animal (n=12 for Lck–Cre Ctcf^f/f mice, n=7 for Lck–Cre Ctcf^+/f mice, and n=14 for wild-type (WT) mice). Note that different scales are used in the vertical axes. Horizontal lines indicate average values (see Table II for actual values±s.e.m.). Lck–Cre Ctcf^f/f mice show increased numbers of DN3 (P<0.01), DN4 (P=0.0002) and ISP cells (P<0.002) compared with wild type. In Lck–Cre Ctcf^f/f mice and heterozygous Lck–Cre Ctcf^f/+ mice DP, CD4 SP and CD8 SP subsets in the thymus were significantly reduced (P<0.0001). CD4 and CD8 T cells in the spleen were significantly reduced in Lck–Cre Ctcf^f/f mice (P<0.00001) and in heterozygous Lck–Cre Ctcf^f/+ mice (P<0.01). (C) Analysis of cell numbers. Ratios were calculated (DP to ISP, CD4⁺ to DP, and CD8⁺ SP to DP) in Lck–Cre Ctcf^f/f mice (hom), Lck–Cre Ctcf^+/f mice (het) and wild-type (WT) mice. The DP/ISP ratio indicates the fold increase in cells and correlates with number of cell divisions. By contrast, the CD4/DP and CD8/DP ratios show a decrease in cell number and are a measure of selection.

In αβ T cells of Lck–Cre Ctcf^f/f mice, a decrease in the proportion and number of DP and SP cells, and a concomitant increase in DN3, DN4 and ISP cells, is observed (Figure 2A and B, Table II). Thus, the effect of a complete Ctcf knockout is particularly prominent at the ISP-to-DP transition (Figure 2B). The fact that the absolute number of Ctcf knockout thymocytes increases about four-fold from ISP to DP, compared with a 90-fold increase in wild-type mice (Figure 2C and Table II), suggests that a cell cycle block rather than increased apoptosis underlies the accumulation of ISP cells from Lck–Cre Ctcf^f/f mice.

Heterozygous Lck–Cre Ctcf^+/f mice also display a phenotype, which is most obvious at the DP stage (Figure 2B, Table II). In these mice, thymic cellularity is only modestly reduced and no accumulation of ISP cells is detected (Table II). These results show that normal CTCF levels are important for proper T-cell development. In agreement with impaired thymic SP cell production, the number of mature CD4⁺ and CD8⁺ T cells in spleen and lymph nodes of Lck–Cre Ctcf^f/f and Lck–Cre Ctcf^+/f mice are significantly reduced (Figure 2A, B; and H Heath, CR de Almeida, RW Hendriks and N Galjart, unpublished data). Interestingly, the ratio between mature CD4⁺ SP and DP cells, and between CD8⁺ SP and DP thymocytes, is similar in heterozygous Lck–Cre Ctcf^+/f, homozygous Lck–Cre Ctcf^f/f and wild-type mice (Figure 2C), indicating that differentiation from the DP to SP stage is not severely affected by a deletion of CTCF.

The accumulation of CTCF-deficient ISP cells in Lck–Cre Ctcf^f/f mice could result from a developmental arrest at the ISP stage or alternatively reflect defective upregulation of CD4 expression in cells that otherwise have characteristics of DP cells, similar to thymocytes deficient for the chromatin remodeler Mi-2β (Williams et al, 2004). To distinguish between these possibilities, we assessed expression of various cell surface markers in wild-type and Lck–Cre Ctcf^f/f thymocytes. Normally, surface expression of TCRβ and its associated signalling molecule CD3 are low in ISP cells and are induced at the DP stage. CTCF-deficient ISP cells express low levels of CD3 and TCRβ (Figure 3A). Furthermore, wild-type and CTCF-negative ISP and DP cells express similar amounts of the surface glycoprotein CD5, which is normally upregulated on ISP cells and somewhat further on DP cells (Azzam et al, 1998). CTCF-deficient ISP cells also express low levels of CD69, which in wild-type mice is induced in a subfraction of DP cells, reflecting TCR-mediated activation (Bendelac et al, 1992). Expression of heat stable antigen (HSA, also called CD24, a cell adhesion molecule), which is normally high on DN and ISP cells and downregulated at the ISP-to-DP transition (Williams et al, 2004), is reduced in CTCF-deficient cells throughout thymocyte differentiation (Figure 3A). As none of the markers tested show a ‘DP-like' expression pattern in CD3-CD4⁻CD8+ cells in Lck–Cre Ctcf^f/f mice, we conclude that these accumulating cells truly represent ISP cells and not abberant DP cells that fail to upregulate CD4 expression. In addition to the ISP-to-DP arrest in Lck–Cre Ctcf^f/f mice, we observed a significant increase in number of cells within the preceding DN3 and DN4 stages (Table II), indicating that a deletion of CTCF causes a developmental arrest from DN3 to ISP.

Figure 3.

Characterization of CTCF-deleted thymocytes. (A) Expression of ISP cell markers. Flow cytometric analyses of HSA, CD3, TCR, CD5 and CD69 in thymocyte subpopulations, displayed as overlays of wild-type mice (black histograms) and Lck–Cre Ctcf^f/f mice (red histograms). Data shown are representative of 5–8 mice per group. (B) Intracellular TCRβ levels in wild-type and Lck–Cre Ctcf^f/f mice. Representative flow cytometric analyses of intracellular TCRβ protein expression in the indicated thymic subsets from a wild-type (WT) and an Lck–Cre Ctcf^f/f mouse. TCRβ/forward scatter (FSC) profiles are shown as dot plots, gating is indicated by the horizontal lines and the percentages of TCRβ⁺ cells are shown. (C) Cell size in wild-type and Lck–Cre Ctcf^f/f thymocytes. Quantification of median forward scatter values (which accurately reflects cell size) of the indicated thymocyte subpopulations in wild-type (grey bars) and Lck–Cre Ctcf^f/f (red bars) mice. Data are average values±s.e.m. from 5 to 8 mice per group.

Interestingly, the expression level of CD3 is not only reduced in ISP thymocytes from Lck–Cre Ctcf^f/f mice, but also in DP cells and—to a lesser extent—in mature CD4⁺ SP and CD8⁺ SP cells (Supplementary Figure S3A, note that in CD8 SP cells, the decrease is not significant). In particular, CTCF-deficient DP cells do not contain the fraction of CD3^high/TCR^high cells. DP cells from heterozygous Lck–Cre Ctcf^+/f mice express CD3 levels that are in the range of those from WT mice (Supplementary Figure S3A), showing that only a complete lack of CTCF affects CD3 expression at the DP cell stage. Although CTCF-negative DP thymocytes are affected in the expression of important molecular markers, including CD3 and TCR, they are still able to differentiate towards the mature SP stages.

Defective cell-size regulation in the TCRαβ lineage in Lck–Cre Ctcf^f/f mice

As our findings indicated a specific function of CTCF at the ISP-to-DP transition, we focussed our attention on possible mechanisms underlying the hampered differentiation of these cells. We first analysed Tcr rearrangement, because of the many CTCF-binding sites that are found in the genes encoding the different receptors. Tcrβ gene rearrangement is generally initiated and completed in DN3. This stage consists of early small DN3E cells that have not yet productively rearranged the Tcrβ locus, and more mature large proliferating DN3L cells expressing TCRβ (Hoffman et al, 1996). We detect a significant increase in the population of large DN2 cells in Lck–Cre Ctcf^f/f mice that contain intracellular TCRβ⁺ (Figure 3B, 13±4% in CTCF-deficient DN2 cells (n=3), compared with 4±0.5% in wild-type controls (n=4); P<0.01). These results raise the interesting possibility that CTCF is involved in inhibiting recombination early in development. The proportion of large TCRβ⁺ DN3 cells appears still elevated, but the difference between the two groups (∼21 and ∼18% in CTCF-deficient and wild-type mice, respectively) is not significant. Thus, the Tcrβ locus can undergo functional V(D)J recombination in cells that have deleted the Ctcf gene. It should be noted that even though the Ctcf gene is efficiently deleted using the Lck–Cre transgene, residual CTCF protein might still be present at DN2 and DN3 stages, and we cannot rule out that this is sufficient for recombination.

By comparing wild-type and Lck–Cre Ctcf^f/f mice, we found that CTCF-negative TCRβ⁺ DN3 cells have the capacity to increase their cell size at the developmental progression of TCRβ− DN3E to TCRβ⁺ DN3L cells (Figure 3B and C). However, from the DN3L-to-ISP stage, CTCF-deficient cells become much smaller than wild-type cells (Figure 3B and C). CTCF-deficient DP cells are again larger than wild-type cells (Figure 3C, Supplementary Figure S3B) and continue to be larger throughout subsequent differentiation (Supplementary Figure S3B). These results indicate that CTCF is a critical regulator of cell growth of αβ T cells. Cells from heterozygous Lck–Cre Ctcf^+/f mice are sized like wild types (Supplementary Figure S3A), showing that only a complete lack of CTCF affects cell size.

Developmental arrest of CTCF-deficient thymocytes is not due to defective Tcr rearrangements

The severe reduction of DP cell numbers and low-surface CD3/TCR expression on CTCF-deficient DP cells, together with the reported presence of CTCF-binding sites in the Tcrα gene locus (Garrett et al, 2005; Kim et al, 2007), indicated that defective TCRα V(D)J recombination could contribute to the arrest of CTCF-deficient thymocytes. We therefore crossed Lck–Cre Ctcf^f/f mice with transgenic mice expressing either the pre-rearranged OTII TCRαβ, which recognizes the OVA_323–339 peptide in the context of C57BL/6 MHC class II, and which positively selects thymocytes towards the CD4 lineage (Barnden et al, 1998), or the MHC class I-restricted HY TCRαβ, which recognizes a male-specific HY antigen peptide in the C57BL/6 H-2^b class I female background and normally drives thymocytes into the CD8 lineage (Kisielow et al, 1988). However, the impaired developmental progression of CTCF-deficient cells is not rescued (Figure 4). Rather, the presence of the OTII and HY TCR transgenes results in an even more severe arrest of T-cell development in the thymus (Figure 4A and B, respectively). OTII Tg Lck–Cre Ctcf^f/f mice manifest a relative increase in the proportions of DN and ISP cells in the thymus, whereas the proportions of DP and CD4 SP cells are reduced. In addition, we observe an almost complete absence of mature T cells in the spleen (Figure 4A). HY Tg Lck–Cre Ctcf^f/f mice manifest a severe block at the ISP stage and an almost complete lack of DP cells in the thymus. Those CD8⁺ cells present in CTCF-deficient HY Tg mice are ISP cells rather than mature CD8 SP cells, because the level of HY-specific TCR expression (as detected by the T3.7 antibody) in the CD8⁺ cell fraction is substantially lower, when compared with wild-type CD8 SP cells. In agreement with this, CD4 or CD8 cells in the spleen are strongly reduced (Figure 4B). Thus, providing Lck–Cre Ctcf^f/f mice with a pre-rearranged TCRαβ does not correct the developmental arrest of DP cells, indicating that the developmental block in Ctcf knockout T cells is independent of Tcr gene rearrangement.

Figure 4.

Arrest of CTCF-deficient thymocytes remains despite Tcr gene rearrangement. (A) Representative flow cytometric analyses of cell populations from thymus (upper two panels) and spleen (lower panel) derived from OTII transgenic mice, which are either in a wild-type (WT) or in a Lck–Cre Ctcf^f/f background. Expression profiles of CD4/CD8 and CD3/TCRγδ markers are shown as dot plots, and the percentages of cells within quadrants or gates are given. Thymic cellularity is shown for the individual animals underneath the second panel. In the right hand panels, data for the OTII transgenics are displayed as histogram overlays. The expression profile of total surface TCRβ within the indicated cell populations is shown for Lck–Cre Ctcf^f/f mice (bold lines) on top of profiles of WT littermates (grey-filled histograms). (B) Representative flow cytometric analyses of cell populations from thymus (upper two panels) and spleen (lower panel) derived from HY transgenic mice, which are either in a wild-type (WT) or in a Lck–Cre Ctcf^f/f background. Flow cytometric profiles of CD4/CD8 and CD3/TCRγδ are shown as dot plots; percentages of cells within quadrants or regions and total thymic cell numbers are given. The expression profile of total HY idiotype-specific T3.7 TCR within the indicated cell populations is shown on the right as a histogram overlay of Lck–Cre Ctcf^f/f mice (bold lines) on top of profiles of WT littermates (grey-filled histograms).

We next crossed Lck–Cre Ctcf^f/f mice with mice deficient for the Recombination activating gene 2 (Rag2) gene. RAG2 mediates V(D)J recombination of Tcr and Ig loci. Rag2^−/− mice are therefore deficient in Tcr gene rearrangement, and they normally do not progress beyond the DN3 stage (Figure 5). In vivo stimulation with anti-CD3ɛ antibodies mimics pre-TCR signalling (Azzam et al, 1998) and thereby induces the formation of DP cells in the absence of rearrangement. Both in Rag2^−/− and in CTCF-deficient Rag2^−/− mice, we find that on anti-CD3ɛ treatment, total thymic cellularity increases from <2 × 10⁶ to 13±4 and 11±3 × 10⁶, respectively (n=4 in each group) and equal numbers of DN4 cells are induced (Figure 5). However, the formation of DP cells is reduced in CTCF-deficient Rag2^−/− mice, when compared with CTCF-expressing cells (Figure 5).

Figure 5.

Arrest of CTCF-deficient thymocytes is independent of Tcrα gene rearrangement. Representative flow cytometric analyses of the thymus of Rag2 knockout (Rag2^−/−) mice, which are either in a wild-type (WT) or in a Lck–Cre Ctcf^f/f background. Mice were either untreated or injected with 50 μg of rat anti-CD3 antibodies (αCD3) in vivo. The CD4 and CD8 expression profiles, 3 days after injection, are shown (upper part). Cell populations were gated (vertical and horizontal lines) and the CD4⁻CD8⁻ fraction (i.e., DN cells) was analysed for CD25 and CD44 expression (lower part). Data are shown as dot plots and the percentages of cells within the quadrants are given. The plots shown are representative for four mice in each group.

As the Tcrγ and δ loci can also undergo functional V(D)J recombination in the absence of CTCF, we conclude that the multiple CTCF-binding sites reported to be present in Tcr loci (Magdinier et al, 2004; Garrett et al, 2005; Barski et al, 2007) do not appear to be essential for the process of V(D)J recombination. Therefore, deficiency of CTCF results in a developmental block at the ISP-to-DP transition that is independent of Tcr gene rearrangement.

Cell cycle arrest during TCRαβ lineage development in Lck–Cre Ctcf^f/f mice

To further examine the underlying cause of the accumulation of CTCF-negative ISP cells, we sorted wild-type and CTCF-deleted cells T cells into DN, ISP and DP fractions, and used real-time PCR to analyse mRNA expression patterns of a selected number of important T-cell factors. In wild-type cells, Ctcf mRNA levels increase from the DN-to-ISP stage and then decrease again in DP cells (Figure 6A). These data suggest that the relatively large, actively cycling ISP cells require a higher amount of CTCF for normal functioning than DN or DP cells. In sorted cells from Lck–Cre Ctcf^f/f mice, Ctcf expression is severely reduced in the DN fraction and is absent at later stages (Figure 6A; note that residual Ctcf mRNA is detected in the DN pool because the Ctcf gene is only fully deleted from DN2 onwards).

Figure 6.

CTCF is important for cell cycle progression. (A) Quantitative RT–PCR analysis in sorted DN, ISP and DP cell fractions from wild-type (WT) and Lck–Cre Ctcf^f/f mice. The DP fraction also contained CD4 SP cells. To obtain enough material, RNA was pooled from four WT and two Lck–Cre Ctcf^f/f mice. (B) Cell cycle status of DN, ISP and DP cells isolated from WT and Lck–Cre Ctcf^f/f mice. Cell cycle was analysed using 7-AAD, which measures DNA amount. A representative analysis is shown. Numbers indicate the percentage of cells in S/G2/M. In Lck–Cre Ctcf^f/f mice, this number is significantly reduced in ISP cells, showing that this population is hampered in the cell cycle. (C) Quantitative RT–PCR (left hand panel) and ChIP (right hand panel) analysis in wild-type and Ctcf^f/f MEFs, after treatment with Cre recombinase. mRNA expression in Cre-treated Ctcf^f/f MEFs (KO) is shown relative to wild type. Although residual Ctcf mRNA is present in the MEFs, p21 expression is increased. ChIP analysis was performed with anti-CTCF antibodies on four regions in the p21 gene. Potential CTCF-binding sites in the mouse p21 gene were chosen on the basis of a genome-wide analysis in human cells (Barski et al, 2007). An identical binding pattern was observed in wild-type MEFs, with relatively weak CTCF binding 2.3 kb upstream of the p21 promoter (up) and on the promoter (pr), and very strong binding on two adjacent regions within intron 1 (i-1a, i-1b). (D) Quantitative RT–PCR (left hand panels) and ChIP (right hand panel) analysis in wild-type thymocytes and T cells before (d0) and after (d3) 3 days of anti-CD3/CD28 stimulation in vitro. mRNA expression of p21 and Ctcf is shown relative to Hprt. ChIP analysis was performed with anti-CTCF antibodies on intron 1 of the p21 gene and, as positive control, on known binding sites in the c-Myc and Tcrα genes.

CTCF was reported to be a negative transcriptional regulator of c-Myc (Lobanenkov et al, 1990; Qi et al, 2003), and c-Myc is important for T-cell development (Trumpp et al, 2001). In wild-type T cells, c-Myc is expressed in a pattern different from Ctcf (Figure 6A). In addition, c-Myc expression is hardly affected by a CTCF deletion. Therefore, in T cells, CTCF does not appear to regulate expression of the c-Myc gene. The level of DNMT1, a maintenance methyltransferase that is essential for T-cell development (Lee et al, 2001), is also similar in wild-type and CTCF-deleted T cells (Figure 6A, see also Figure 1C). Two other transcription factors are GATA3, which is critically involved in β-selection and development of CD4 SP cells (Pai et al, 2003), and SATB1, which organizes cell type-specific nuclear architecture (Cai et al, 2006). In wild-type cells, the levels of these factors are opposite to those of Ctcf (Figure 6A). In CTCF-deficient cells, Gata3 and Satb1 expression is reduced, in particular at the DP stage. We next tested two cytoplasmic factors involved in T-cell signalling. Expression of PreTα (that assembles with TCRβ to initiate TCR signalling) is not affected in ISP cells and is up- rather than downregulated in DP cells in the absence of CTCF (Figure 6A). Finally, the expression of GIMAP4, which is induced by pre-TCR signalling and accelerates T-cell death (Schnell et al, 2006), is increased in CTCF-deficient DP T cells with a factor of ∼2 (Figure 6A). Nevertheless, we did not find evidence for increased apoptosis of CTCF-deficient DP cells (data not shown). Although none of the factors tested appears to be directly regulated by CTCF, changes in their expression level may contribute to the observed phenotype in CTCF-deleted cells.

As the accumulation of ISP cells from Lck–Cre Ctcf^f/f mice could be due to cell cycle defects, we tested the expression of two major cell cycle inhibitors, p21 and p27. In wild-type cells, the expression profile of these factors is opposite to that of Ctcf, whereas Ctcf knockout cells show significantly increased p21 and p27 expression (Figure 6A). We subsequently analysed cell cycle profiles in wild-type and CTCF-deleted thymocytes. The CTCF-deficient ISP population from Lck–Cre Ctcf^f/f mice contains approximately half the number of cycling cells compared with wild type (29%±1 cells in S/G2/M phase in Lck–Cre Ctcf^f/f mice (n=3), versus 53%±8 in wild-type mice (n=3); Figure 6B shows an example of analysis from individual mice). These results indicate that cell cycle progression in β-selected CTCF-deficient T cells is blocked due to the upregulation of p21 and p27.

We next tested whether a deletion of CTCF always causes upregulation of the p21 and p27 genes, irrespective of cell type. We treated mouse embryonic fibroblasts (MEFs) from Ctcf^f/f mice with Cre recombinase (Splinter et al, 2006) to efficiently remove CTCF in these cells. Real-time PCR experiments show that in MEFs the expression of p21, but not of p27, is increased in the absence of CTCF (Figure 6C). Thus, the combined upregulation of p21 and p27 seen in CTCF-deficient ISP cells is not a general regulatory mechanism.

CTCF as a critical regulator of cell growth and proliferation in T cells

Genome-wide studies in human cells have shown that CTCF-binding patterns in fibroblasts and T cells largely overlap (Barski et al, 2007; Kim et al, 2007). Interestingly, the human p21 gene contains four CTCF binding sites in the vicinity of its promoter, whereas no binding sites are found near p27. As p21 but not p27 mRNA is upregulated in different cell types in the absence of CTCF, and CTCF binds near and within the human p21 gene but not near p27, we hypothesized that CTCF might be a negative regulator of p21 expression. This idea is enforced by the observation that in wild-type DN, ISP and DP thymocytes, the p21 mRNA expression profile is exactly opposite to that of Ctcf (Figure 6A). To test our assumption, we first confirmed that CTCF-binding sites (which are part of CNEs) are conserved between man and mouse. Chromatin immunoprecipiations (ChIPs) in MEFs on the corresponding regions of mouse p21 indeed reveal an identical pattern of CTCF binding, including the two adjacent strong sites in intron 1 (Figure 6C).

Next, we analysed thymocytes isolated from total thymus and largely consisting of DP cells, and purified CD4⁺ T cells, either before (day 0) or after 3 days of in vitro activation by anti-CD3/anti-CD28 stimulation (day 3). We reasoned that in these cells, levels of p21 must differ and that these cells therefore represent a good model to test whether CTCF binding correlates to p21 expression in cells expressing normal amounts of CTCF. We correlated mRNA expression levels of Ctcf and p21 (as tested with real-time PCR) with CTCF binding (as analysed by ChIP), using the strongest CTCF site in intron 1 of the p21 gene as a reference. mRNA expression levels were normalized to Hprt (a housekeeping gene), whereas CTCF binding was normalized to the Amylase gene, which contains no CTCF-binding sites. The data show that CTCF binds very strongly to the p21 intron, both in thymocytes and resting and cultured T cells (Figure 6D). Although there is a correlation between Ctcf mRNA levels and strength of CTCF binding, there is no correlation with p21 expression (Figure 6D). Thus, despite the fact that deletion of CTCF results in increased expression of p21 in thymocytes and MEFs, the p21 gene does not appear to be a general target of CTCF in wild-type cells.

The mRNA expression data indicate that Ctcf is specifically upregulated in ISP cells (Figure 6A). Furthermore, lacZ staining results show increased activity of the Ctcf promoter in DN2-DN4 cells (Figure 1D) and in ISP cells (not shown). However, both the PCR and LacZ staining results reflect mRNA levels of CTCF. To also analyse CTCF protein levels in vivo, we used a Ctcf^gfp knock-in allele in which GFP–CTCF is expressed instead of CTCF (H Heath and N Galjart, unpublished data). We used flow cytometry to identify GFP–CTCF in the different thymocyte subsets and in the spleen (Supplementary Figure S4). During αβ T-cell differentiation, CTCF levels increase from the DN2 to the ISP stage and then return to DN1 values in the DP cell population (Supplementary Figure S4A), consistent with the expression profile of Ctcf transcripts (Figure 6A). In TCRαβ CD4⁺ and CD8⁺ T cells in the spleen, CTCF levels are identical to the levels in DN1 and SP cells in the thymus (Supplementary Figure S4B). Thus, CTCF levels increase in those thymocyte populations (DN3, DN4 and ISP) that have a larger cell size (see Figure 3). Interestingly, a lack of CTCF causes accumulation of these very same cells. Taken together, our data suggest that CTCF is a critical regulator of cell growth and proliferation in αβ T cells.

Discussion

CTCF is an important protein involved in chromatin organization and the epigenetic regulation of gene expression. Most studies on CTCF use cultured cells as a basis, and a plethora of results regarding CTCF function have been reported. We have generated a conditional Ctcf knockout allele, which has allowed us to examine the in vivo function of CTCF. Consistent with a previous report (Fedoriw et al, 2004), we find that deletion of Ctcf in early embryonic development is lethal. A novel result is that heterozygous Ctcf knockout mice, which are viable and fertile, are born in less than expected numbers. The development of Ctcf^+/− thymocytes is also affected, although mildly. Taken together, these data suggest that CTCF is required in a dose-dependent manner.

We found no evidence for an increased tumour incidence in heterozygous Ctcf knockout animals, or for T lymphoid malignancies in CTCF-deficient T-cell lineages. This argues against a function of CTCF as a crucial tumour suppressor (Klenova et al, 2002). Increased expression of p21 (and p27) in CTCF-negative cells would explain why loss of CTCF does not induce tumours. Still, the potential function of CTCF in cancer merits a more detailed investigation. We also do not observe a more severe phenotype in CTCF-deficient female thymocytes compared with male cells, suggesting that absence of CTCF does not cause deregulated expression of genes on the inactive X-chromosome. Moreover, the deletion of CTCF does not have an effect on the maintenance of methylation in the imprint control region of the Igf2/H19 and ribosomal DNA (rDNA) loci (see Supplementary data). Unlike other studies (Schoenherr et al, 2003; Filippova et al, 2005), our results therefore indicate that CTCF is not required to maintain X-inactivation and DNA methylation status of the Igf2/H19 locus.

Despite the presence of multiple CTCF-binding sites near the Tcrα and Tcrβ genes, our data indicate that CTCF is not essential for recombination at these loci. CTCF also does not appear to be required for the differentiation of DP cells towards the CD4⁺ and CD8⁺ SP cell stage, even though substantial epigenetic and regulatory changes accompany commitment of SP cells (Rothenberg and Taghon, 2005). CTCF is, however, essential for the efficient proliferation of β-selected cells, in particular for their maturation from ISP to DP cells, and for TCR upregulation at the cell surface of DP cells. As T cells were directly isolated from mice, our data provide the first in vivo evidence for an important function of CTCF in cell cycle progression. In line with a proliferative block, we detect strongly increased expression of two major cell cycle inhibitors, p21 and p27.

Our results in T cells are completely opposite to those obtained in WEHI 231 B lymphoma cells, where conditional expression of CTCF resulted in the up- rather than the downregulation of p21 and p27, whereas reduction of CTCF levels decreased rather than increased the expression of p21 and p27 (Qi et al, 2003). This could be due to the fact that the properties of the ISP thymocytes and WEHI 231 B cells are entirely different. ISP thymocytes are highly proliferating as a result of pre-TCR stimulation, whereas crosslinking of the B-cell receptor on WEHI 231 immature B cells results in cell-cycle arrest and apoptosis (thereby providing a model for self-tolerance by clonal deletion). This suggests that CTCF function is context dependent, although it should be noted that our data were obtained in vivo, whereas the data in the WEHI 231 B lymphoma cells were obtained with stably transfected clones selected for high expression of CTCF sense or antisense mRNA (Qi et al, 2003). Using our conditional Ctcf^f/f mice in combination with existing B cell-specific Cre transgenes, we will be able to examine the function of CTCF in B cells.

In Ctcf knockout mice, approximately four times more DP cells are present than ISP cells, suggesting that the latter cells can divide, although slowly. Moreover, whereas CTCF-negative ISP cells are smaller than their wild-type counterparts, DP cells are larger. These data argue against increased apoptosis as a cause of reduction in DP cell numbers after Ctcf gene deletion. This is different from the situation in Dnmt1 knockout mice, which have similar numbers of DN thymocyte subsets as wild-type littermates, and in which increased apoptosis was shown to be a major cause of the severe reduction in DP cell number (Lee et al, 2001).

T-cell activation at the DN3 stage is accompanied by an enlargement of both cytoplasmic and nuclear volume. In the DN4 and ISP stages, growth is coupled to proliferation, presumably to sustain the rapid cell divisions that are required in these cells. When thymocytes enter the DP stage, they exit the cell cycle and become small again. CTCF levels increase when cells become bigger (from DN3 to ISP) and decrease again when cells become small. Furthermore, CTCF-negative DN4 and ISP cells are smaller than wild-type cells, whereas DP cells are larger. These data uncover a function of CTCF as a regulator of cell growth and proliferation in thymocytes. A notable enrichment of CTCF-binding sites was observed near DNAse I HSs in CD4⁺ T cells, indicating that CTCF controls global T-cell expression (Boyle et al, 2008). Our data suggest that CTCF couples cell growth to the cell cycle. Strikingly, the ratio between DP cell number and mature CD4⁺ SP and CD8⁺ SP thymocytes is the same in wild-type, heterozygous Lck–Cre Ctcf^+/f and homozygous Lck–Cre Ctcf^f/f mice. Thus, the major function of CTCF in the thymus might be to positively regulate cell growth in rapidly dividing thymocytes so that appropriate numbers of cells are generated before positive and negative selection events at the DP stage in the thymus.

Materials and methods

Modified Ctcf alleles, mouse models and embryonic fibroblasts

Human CTCF cDNA was used to screen a 129S6/SvevTac mouse PAC library (RPCI-21) (Osoegawa et al, 2000). PAC clones were used to isolate 6.7 kb (for 5′-end targeting) and 8 kb (for 3′-end targeting) EcoRI subclones. For 5′-end targeting, the 6.7 kb EcoRI fragment was used to amplify 1360 bp of 5′-end homology and 5340 bp of 3′-end homology. The homologous arms were cloned into a vector containing the neomycin resistance gene flanked by loxP sites (Hoogenraad et al, 2002). A viral thymidine kinase gene was inserted afterwards. For 3′-end targeting, we generated a SpeI–EcoRI subclone from the PAC DNA and used its unique BamHI site to insert a cassette containing the puromycin resistance gene flanked by loxP sites, followed by splice acceptor sequences and the bacterial β-galactosidase (lacZ) reporter (Hoogenraad et al, 2002). Relevant parts of the different constructs were verified by DNA sequencing.

Constructs were targeted into E14 embryonic stem (ES) cells as described (Hoogenraad et al, 2002). DNA from resistant ES cells was analysed with external radiolabelled probes by Southern blotting. Confirmation of homologous recombination was performed using different 5′-end and 3′-end probes (Figure 1A and B) and a PCR-based assay for genotyping. Ctcf^f/f mice were crossed back more than 10 times to the C57BL/6 background. MEFs were isolated from Ctcf^f/f mice using published procedures (Akhmanova et al, 2005). Fibroblasts were treated with lentiviral Cre as described (Splinter et al, 2006).

Ctcf^f/f mice were bred to mice expressing chicken β-actin-Cre-generating Ctcf^+/− animals. T-cell-specific deletion of Ctcf was achieved by breeding to Lck–Cre mice (Lee et al, 2001), which were kindly provided by Dr C Wilson (University of Washington, Seattle, WA, USA). Cre-specific primers were used for genotyping. In experiments where Ctcf knockout mice are compared with ‘wild type' animals, the latter are littermates of the Lck–Cre Ctcf^f/f animals, that is, mice that express normal (‘wild type') levels of CTCF. Thus, the ‘wild type' mice may be truly wild type; they may contain the Lck–Cre transgene or the Ctcf^f allele, but never a combination of the latter two alleles.

HY/Rag2^−/− (C57BL/10) mice were purchased from Taconic Europe A/S (Denmark). OT-II mice have been described (Barnden et al, 1998). Mice were bred and maintained in the Erasmus MC animal care facility under specific pathogen-free conditions and analysed at 6–10 weeks of age. For anti-CD3 treatment, Rag2-deficient mice (Shinkai et al, 1992) were injected i.p. with 50 μg of rat anti-CD3 antibodies (αCD3; clone 145-2C11) as described (Levelt et al, 1995). Experimental procedures were reviewed and approved by the Erasmus University committee of animal experiments.

DNA, RNA and protein analysis

Genomic DNA was isolated, digested and blotted onto Hybond N+ membranes (Amersham) and hybridized with radiolabelled probes. Ctcf probes are shown in Figure 1. Total RNA was prepared using RNA-Bee RNA isolation solvent (Tel-Test Inc.). RNA (0.5–1.0 μg) was reverse-transcribed (RT) with random and oligo-dT primers, in the presence of Superscript reverse transcriptase (Invitrogen). For the experiment shown in Figure 6, RNA was isolated and pooled from the thymus of four wild-type mice and from two Lck–Cre Ctcf^f/f mice.

Real-time RT–PCR was performed as described (Splinter et al, 2006) with 100 ng of each primer and 0.5 U of Platinum Taq DNA polymerase (Invitrogen). Sybr-green (Sigma) was added to the reactions and PCR was performed on a DNA Engine Opticon PCR system (MJ Research Inc.) and Bio-Rad MyiQ iCycler single-colour real-time PCR detection system. To confirm the specificity of the amplification products, samples were separated by standard agarose gel electrophoresis. Threshold levels were set and further analysis was performed using the SDS v1.9 software (Applied Biosystems). The obtained C_t values were normalized to the C_t value of Gapdh, β-actin or Hprt. Each PCR was performed at least in triplicate, and at least two independent experiments were performed to examine the expression of individual genes. Primer sequences and PCR conditions used are available on request.

For ChIP, nuclear extracts were prepared (Splinter et al, 2006). Chromatin cross-linking (2 × 10⁷ cells treated with 1% formaldehyde for 10 min at room temperature), sonication to 300–800 base pair fragments and immunoprecipitation were as described (Upstate protocol, http://www.upstate.com). At least two independent ChIPs were carried out per experiment. Quantitative real-time PCR was performed as described above. Values were normalized to input measurements and enrichment was calculated relative to the Amylase gene using the comparative C_t method. PCR products were all smaller than 150 bp. Primer sequences and PCR conditions used are available on request.

Western blot analysis was performed as described (Hoogenraad et al, 2002). Primary antibody incubation was done overnight at 4°C in Tris-buffered saline, containing 5% (w/v) BSA and 0.15% (v/v) NP-40. Blots were incubated with secondary goat anti-rabbit or anti-mouse antibodies, coupled to horseradish peroxidase (GE Healthcare UK Ltd: 1:50 000). Signal detection was performed using ECL (Amersham). Anti-CTCF (N3) and anti-fibrillarin (no. 4118 and 4080) antibodies were generated as described (Hoogenraad et al, 2000) using GST-linked chicken CTCF (amino acids 2–267) and fibrillarin fusion proteins. These antisera were used in a 1:300 dilution. DNMT1 (Abcam), and UBF (Santa Cruz Biotechnology) mAbs were used at 1:100. Western blots were scanned and the levels of protein were quantified using the gel macro function in ImageJ (Rasband, WS, NIH, http://rsb.info.nih.gov/ij/). The amount of CTCF was normalized to DNMT1 in the same sample.

Flow cytometric analyses

Preparation of single-cell suspensions, FDG-loading, mAb incubations for four-color cytometry have been described (Hendriks et al, 1996). All mAbs were purchased from BD Biosciences (San Diego, CA). Samples were acquired on a FACSCalibur™ flow cytometer and data was analysed using CellQuest™ software (BD Bioscience). For cell cycle profiles of thymic subsets, cells were first stained for surface markers, fixed with 0.25% paraformaldehyde and permeabilized with 0.2% Tween 20. Next, 7-AAD was added to a final concentration of 15 μg/ml in PBS. Cell cycle status of T-cell cultures was determined after fixing in ice-cold ethanol and subsequent staining in PBS, containing 0.02 mg/ml propidium iodide, 0.1% v/v Triton X-100 and 0.2 mg/ml RNAse. Doublet cells were excluded by measuring peak area and width. FACS sorting of DN, ISP and DP cells was performed with a FACSVantage VE equipped with Diva Option and BD FACSdiva software (BD bioscience). The purity of fractions was >98%.

Statistical analysis

Statistical evaluations were performed by standard two-tailed t-test.

Supplementary Material

Supplementary Information