A DNA fragment harboring C to A nucleotide change that generates Bcl11b^N797K protein was generated by overlap PCR, and replaced with wt exon 4 in the target vector, which was used to generate the Bcl11b^{flo x} allele (9). The targeting vector for Bcl11b^N797K mutation was transfected into M1 embryonic stem (ES) cells as previously described (9), and ES clones which underwent homologous recombination were identified by PCR with an appropriate primer set. Bcl11b^+/N797K mice were generated by crossing C57BL/6NJcl mice purchased from CLEA Japan (Shizuoka, Japan) with chimeric mice generated from ES clones with Bcl11b^+/N797K genotype. The Bcl11b<σπ>Δ</σπ> allele was generated from the Bcl11b^fl allele (9) by crossing Bcl11b^+/fl mice with EIIa-Cre transgenic mice that were purchased from Jackson laboratory (#003314). Cd4^ΔS mice was previously described (24). CD45.1 congenic mice and Rag1-deficient mice were obtained from Jackson Laboratories. All mice were maintained in the animal facility at the RIKEN IMS. All animal procedures were in accordance with institutional guidelines for animal care and the protocol (2020–026) approved by the Institutional Animal Care and Use Committee of RIKEN Yokohama Branch. All the mice were sacrificed by euthanasia with CO₂ overdose following anesthesia using isoflurane.

Thymus, spleen, lymph nodes and other organs were harvested from mice at various, mashed and passed through 100 μm pore cell strainer to make single cell suspensions. After hemolysis with ACK Lysing Buffer (Thermo Fisher Scientific, MA), cells were washed with ice-cold staining buffer (D-PBS (-), 2mM EDTA, 0.05% NaN₃, and 2% FBS) followed by blocking with 2.5 μg/ml of 2.4G2 antibody (BD Biosciences, NJ). The small intestine was dissected, and intestinal contents were washed out with cold RPMI-1640 medium (SIGMA). The intestine was cut into approximately 1.0 cm pieces and incubated in the RPMI-1640 medium containing 5 mM EDTA and 2% fetal bovine serum (FBS) for 15 min at 37°C, followed by another cycle of incubation in RPMI-1640 medium with 2% FBS. Remaining intestinal tissues were cut into small pieces and digested with 1.0 mg/ml collagenase (SIGMA) suspended in RPMI-1640 medium for 15 min at 37°C. The resultant supernatants from the collagenase digestion were collected and passed through a 100 μm cell strainer after 3 cycles of these steps. The cells were subjected to Percoll (G.E. Healthcare) gradient separation and lymphocytes in the interphase were collected and proceeded for the flow cytometric analysis.

Surface molecules were stained with specific antibodies by incubating for 30 min on ice. The following antibodies for surface molecules were purchased from BD Biosciences or Thermo Fisher Scientific; B220 (clone: RA3-6B2), CD3ε (clone: 145-2C11), CD4 (clone: RM4-5), CD8α (clone: 53-6.7), CD11b (clone: M1/70), CD19 (clone: 1D3), CD24 (clone: M1/69), CD25 (clone: PC61.5), CD44 (clone: IM7),CD45 (clone: 30-F11, or clone:104), CD45.1 (clone: A20), CD45.2 (clone: 104), CD62L (clone: MEL-14), CD127 (clone:A7R34), c-Kit (clone: 2B8), Gr-1 (clone: RB6-8C5), IL-7Rα (clone: A7R34), ST2 (clone: DIH4), Sca1 (clone:E13-161.7), NK1.1 (clone: PK136), TCRβ (clone: H57-597) and γδTCR (clone: GL3). α-galactosyl ceramide (a-GalCer) loaded CD1d tetramer was prepared by mixing recombinant dimeric mouse CD1d:Ig fusion protein (BD Biosciences, DimerX I). In order to investigate intracellular molecules, cells were fixed and permeabilized by use of Transcription Factor Buffer Set (BD Biosciences) following surface staining. Transcription factors were detected with following antibodies purchased from BD Biosciences or abcam (Cambridge, UK); Bcl11b (clone: 25B6), Foxp3 (clone: FJK-16s), Gata3 lone: 16E10A23), Rorγt (clone: Q31-378), and Thpok (clone: T43-94). Dead cells were distinguished by use of 7-amino-actinomycin D (7-AAD) (BD Biosciences) and LIVE/DEAD^® Fixable Dead Cell Stain (Thermo Fisher Scientific) for general and intracellular staining, respectively. Multi-color flow cytometric analysis was performed using BD FACSCanto II™ (BD Biosciences), and data was processed with FlowJo™ software (BD Biosciences). Cell subsets were sorted using a BD FACSAria™ III (BD Biosciences).

Ten million bone marrow cells were isolated from femur of mice at 6 weeks of age with either CD45.2 Bcl11b^+/+ or Bcl11b^+/N797K genotype by flushing, and mixed with equal number of CD45.1 Bcl11b^+/+ bone marrow cells in 200 μl of D-PBS (-). Bone marrow cells were intravenously injected into lethally irradiated (9.5 Gy) Rag1-deficient recipients. Fetal liver (FL) cells isolated from E13.5 or 14.5 Bcl11b^+/+ and Bcl11b^N797K/N797K embryos were mashed and passed through 100 μm strainer to prepare single cell suspension. FL cells were washed with ice-cold D-PBS (-) once, suspended in 200 μl of D-PBS (-), and injected into lethally irradiated (9.5 Gy) Rag1-deficient mice intravenously. Drinking water containing antibiotics (1 μg/ml neomycin and 100 unit/ml polymyxin B) was supplied to recipient mice for the first two weeks post-transplantation. Three months after the transplantation, lymphoid organs were isolated from mice and flow-cytometric analyses were performed.

Frozen fetal liver (FL) samples were thawed and cultured in OP9 media (alpha MEM [Gibco] supplemented with 20% FBS, 50μM β-ME (Sigma), penicillin-streptomycin-glutamine supplement [Gibco]) overnight with 10 ng/ml IL-7 and Flt3L and 5 ng/ml KitL (Peprotech). The following day, cells were depleted by biotinylated antibodies (Biolegend) and anti-biotin magnetic beads (Miltenyi) for lineage markers (TCRβ, CD19, CD11c, NK1.1, Ter119, γδTCR, Gr-1). Remaining Lineage-negative cells were subsequently cultured on an OP9-DLL1 cell monolayer in OP9 medium with IL-7 and Flt3L (10ng/ml until day 5, 5 ng/ml until day 8, 1 ng/ml subsequently) and split every 2-3 days.

On harvesting, a known fraction of the culture was immediately removed, and mixed with a known volume of a resuspended mix of PBS (Gibco), SytoxBlue (Invitrogen) and single colour Calibrite compensation beads (BD). The number of beads in the counting mix was established with a hemocytometer. Bead/Cell mixes were vigorously resuspended by pipetting prior to acquisition by flow cytometry (MACSQuantify, Miltenyi). On flow cytometric analysis (FlowJo software, BD) the ratio of viable (SytoxBlue-negative) and non-OP9 (GFP-negative) cells to beads was established, and the absolute number of cells calculated. The dilution factor of each split was accounted for to calculate absolute cell numbers, and these were normalized to day 5 (as 100%) to account for cell number differences between samples. Population frequencies of total viable cells were extracted by flow cytometry and multiplied by these normalized absolute cell counts.

Total thymocytes were harvested from lethally irradiated Rag1-deficient recipients, in which thymocyte development was reconstituted with Bcl11b^+/+ and Bcl11b^N797K/N797K FL cells at eight weeks after the transplantation, and CD4⁻CD8α⁻CD44⁺CD25⁺ DN2 thymocytes were purified by cell sorting. Cells were seeded on TSt-4/Dll4 feeder cells, and cultivated in RPMI1640 medium supplemented with 10% FBS, 50 μM β-ME, antibiotics, and cytokine cocktail (5 ng/ml each of KitL, Flt3L, and IL-7). Five days after, cells were harvested by vigorous pipetting and were passaged to new feeder cells. Flow cytometric analyses were performed on day 5^th and 9^th. Dead cells and feeder cells were distinguished from hematopoietic cells by 7-AAD staining and CD45 expression. The absolute numbers of hematopoietic cells in culture were calculated, and were normalized to day 0 (as 100%).

One hundred thousand target cells (DN [CD24⁺TCRβ⁻CD4⁻CD8α⁻], ISP [CD24⁺TCRβ⁻CD4⁻CD8α⁺] and DP [CD24⁺TCRβ⁻CD4⁺CD8α⁺] thymocytes) were purified by cell sorting. Cells were washed with ice-cold D-PBS(-), and ATAC libraries were prepared by use of ATAC-Seq Kit (Active Motif, CA) according to manufacturer’s protocol. To deplete large DNA fragments, prepared library DNA solution was mixed with x 0.65 volume of SPRIselect beads (Beckman Coulter, CA) and supernatant was collected. Supernatant was then mixed with x 0.65 volume of beads, and DNA fragments captured on the beads were eluted with 0.1x TE buffer resulted in an enrichment of fragments with 150-500 bp length. Fragment size was validated by Agilent, 2100 Bioanalyzer and High Sensitivity DNA Kit (Agilent, CA). Libraries were quantified by quantitative PCR using KAPA Library Quantification Kit (Roche). Sequencing was performed with Illumina HiSeqX (Illumina, CA) at Macrogen (Seoul, South Korea).

ATAC-sequence reads were trimmed and their quality were checked using trimgalore (version 0.6.7, https://github.com/FelixKrueger/TrimGalore) and then mapped to the mm10 genome using bowtie2 (version 2.4.5, https://bowtie-bio.sourceforge.net/bowtie2/index.shtml). Mapped reads were normalized by total reads and window size and were expressed as RPKM (reads per kilobase per million reads). Statistical analysis was performed by F-test and unpaired Students’ t test with or without Welch’s correction using GraphPad Prism 7 (Graphpad software, CA) or Excel (Microsoft, WA). ATAC-Seq data are available at under accession number SRP473952.

The position of the N807 residue replaced with K in human patients harboring the BCL11B^N807K variant corresponded to the N797 position in the mouse Bcl11b protein ( Supplementary Figure 1A ). Therefore, we introduced a C-to-A single nucleotide mutation via gene targeting using mouse ES cells in the mouse Bcl11b gene, hereafter referred to as the Bcl11b^N797K allele, which produces the Bcl11b^N797K protein ( Supplementary Figure 1B ). A mouse strain harboring the Bcl11b^N797K mutation was established through germline transmission from chimeric mice generated from a Bcl11b^+/N797K ES cell clone ( Supplementary Figure 1C ). We first noticed that heterozygous Bcl11b^+/N797K founders showed slight growth retardation and did not gain normal weight at weaning ( Supplementary Figure 1D ). Bcl11b^+/N797K mice also tended to die at one week of age, and only half of then survived at six weeks of age on a normal diet ( Supplementary Figure 1E ). Our RNA-seq data using thymocytes from Bcl11b^+/N797K mouse confirmed that Bcl11b mRNA was transcribed equally from both the Bcl11b⁺ and Bcl11b^N797K alleles ( Supplementary Figure 1F ), and Bcl11b protein expression in DP thymocytes was comparable between Bcl11b^+/+ and Bcl11b^+/N797K mice ( Supplementary Figure 1G ).

Since the patient with de novo heterozygosity for the BCL11B^N807K variant showed impaired T cell development (16), we first examined T cell development in Bcl11b^+/N797K mice at four weeks of age using flow cytometry. Thymocytes cellularity were reduced to approximately one-tenth in Bcl11b^+/N797K mice compared to Bcl11b^+/+ mice ( Supplementary Figure 2A ). The CD4 and CD8α expression profiles showed an increase in the frequency of the CD4⁻CD8⁺ SP population and a decrease in the frequency of the CD4⁺CD8⁻ SP population ( Figure 1A ). The CD4⁻CD8⁺ SP population consists of immature (CD24⁺TCRβ^lo) and mature (CD24⁻TCRβ^hi) thymocytes. Analyses of CD24 and TCRβ expression revealed that the frequency of the immature CD24⁺TCRβ^lo CD8 SP thymocytes, hereafter referred to as ISP8 thymocytes, was elevated to more than 80% in the thymus of Bcl11b^+/N797K mice ( Figure 1A ). Accumulation of ISP8 thymocytes by the Bcl11b^N797K protein indicates that thymocytes differentiation was impaired by the Bcl11b^N797K mutant at the transition from the CD4⁻CD8⁻ DN to the CD4⁺CD8⁺ DP stage. In the CD4⁻CD8⁻ DN thymocytes population, the frequency of the CD44⁺CD25⁺ DN2 subset was increased with a tendency for a decreased percentage of the CD44⁻CD25⁺ DN3 subset in Bcl11b^+/N797K mice compared to Bcl11b^+/+ mice ( Figure 1B ). We also noticed that the percentage of mature (CD24⁻TCRβ^hi) thymocytes was lower in Bcl11b^+/N797K mice than in Bcl11b^+/+ mice ( Figure 1C ). In addition, the percentages of CD4⁻CD8α⁻ and CD4⁺CD8α⁺ cells in the mature thymocyte population were elevated in Bcl11b^+/N797K mice, as was observed in the mutant mice lacking the last ZnF motif in the Bcl11b protein (9).

Bcl11b is also essential for the development of iNKT (12) and Foxp3⁺ regulatory T (Treg) cells (10), two lineages that are known to be positively agonist-selected (25). The frequency of iNKT cells that were detected as TCRβ⁺ cells bound by αGalCer-loaded CD1d-dimer in the CD24⁻ mature thymocyte population was significantly lower in the thymus of Bcl11b^+/N797K mice than in Bcl11b^+/+ mice ( Figure 1D ). In the thymic iNKT population, there was a decrease in the frequency of NK1.1⁻CD44⁺ stage 2 subset in the Bcl11b^+/N797K mice compared with controls ( Supplementary Figure 2B ). Thus, Bcl11b^N797K mutant interferes with iNKT differentiation at the early developmental stage. In contrast, the frequency of Foxp3⁺ Treg cells in the CD4 SP thymocyte population was higher in Bcl11b^+/N797K mice than in Bcl11b^+/+ mice ( Figure 1E ). These results indicate that the mutant Bcl11b^N797K protein perturbs early T cell development, as expected in a human patient (16). Bcl11b is also required for the development of ILC2s in mice (13, 14), and some human patients harboring other BCL11B variants showed a reduction in ILC2s in their blood (16, 18). We therefore next examined ILC2 development in our mutant mice. The frequency of Gata3⁺Rorγt^- ILC2s among the small intestine CD45⁺CD3ε⁻IL-7Rα⁺ lamina propria cells was modestly, though statistically, lower in Bcl11b^+/N797K than in Bcl11b^+/+ mice ( Supplementary Figure 2C ). Thus, as observed in the reduction of ILC2s in the blood of BCL11B^+/N807K human patients, the corresponding mis-sense mutation Bcl11b^N797K in mice likely impairs ILC2 development.

In the peripheral lymphoid tissues, the total number of splenocytes was less than one-fifth in Bcl11b^+/N797K mice than in Bcl11b^+/+ mice ( Supplementary Figure 2D ), presumably due to their smaller body sizes. As expected, the frequency of TCRβ⁺ T cells in the spleen was decreased in Bcl11b^+/N797K mice ( Supplementary Figure 2D ). Although, the CD4/CD8 ratio was comparable between Bcl11b^+/N797K and Bcl11b^+/+ mice, both CD4⁺ and CD8⁺ splenic T cells tended to differentiate into CD44⁺ memory-phenotype T cells ( Supplementary Figure 2E ). For instance, the frequency of the CD44⁺CD62L⁺CD8⁺ T cell subset (CD8⁺ T central memory; TCM) and CD44⁺CD62L⁻ CD4⁺ T subset (CD4⁺ T effector memory; TEM) and CD44⁺CD62L⁻ CD8⁺ T cell subset (CD8⁺ TEM) were elevated, while that of the CD44⁻CD62L⁺ naïve T cell subset was reduced in Bcl11b^+/N797K mice ( Supplementary Figure 2E ). This increase in memory phenotype T cells has been reported as a BCL11B-related disorder (BCL11B-RD) obseved in human patients (26).

To examine whether T cell development in Bcl11b^+/N797K mice is impaired in a hematopoietic cell-intrinsic manner and the developmental potency of Bcl11b^+/N797K progenitors in a competitive setting against Bcl11b^+/+ cells, we performed mixed bone marrow chimera experiments. A mixture of CD45.1⁺ Bcl11b^+/+ bone marrow cells and either CD45.2⁺ Bcl11b^+/+ or Bcl11b^+/N797K bone marrow cells was injected into lethally irradiated Rag1-deficient mice ( Supplementary Figure 3A ). Eight weeks after transplantation, the ratio of CD45.2⁺ to CD45.1⁺ cells was examined among several types of hematopoietic cells. Compared with the Gr-1⁺ myeloid and B220⁺ B lymphocyte populations in the circulating blood, the frequency of CD45.2⁺ cells in CD3ε⁺ T cells was lower in recipient mice that received CD45.2⁺ Bcl11b^+/N797K bone marrow cells ( Supplementary Figure 3B ). In the spleen, we confirmed that Bcl11b^+/N797K progenitors gave rise to B220⁺ B lymphocytes, whereas the frequency of CD45.2⁺ cells in both CD3ε⁺TCRβ⁺ and CD3ε⁺γδTCR⁺ cells was significantly reduced in recipients that received CD45.2⁺ Bcl11b^+/N797K bone marrow cells ( Figure 2A ). This reduction was more severe in the CD3ε⁺γδTCR⁺ population than in the CD3ε⁺TCRβ⁺ population. Thus, the developmental potency to generate mature T cells in Bcl11b^+/N797K progenitors was nearly outcompeted by the control progenitors under equally mixed settings. Next, we examined at which T developmental stages, Bcl11b^+/N797K progenitors were outcompeted. In the thymus, while the ratio of CD45.2⁺ Bcl11b^+/N797K cells to CD45.1⁺ Bcl11b^+/+ cells was nearly one in the DN1 population, it started to decline from the CD44⁺CD25⁺ DN2a stage and was significantly lower at the CD44⁻CD25⁺ DN3 stage ( Figure 2B ). These results indicated that the Bcl11b^N779K mutant protein may impair the generation of DN2a thymocytes. Interestingly, the frequency of CD45.2⁺ cells increased at the ISP8 stage, followed by a sharp decrease at the DP stage ( Figure 2B ). The accumulation of CD45.2⁺ Bcl11b^+/N797K ISP8 thymocytes relative to the control CD45.1⁺ Bcl11b^+/+ cells, which stems from a developmental block at the transition from the DN4 to DP stage by the Bcl11b^N797K protein, is likely a cause of the increase in the CD45.2^/CD45.1 ratio at the ISP8 stage. Thus, the potency of generating DP thymocytes from Bcl11b^+/N797K progenitors was outcompeted at this transitional stage.

We also examined the ratio of CD45.2⁺ to CD45.1⁺ cells in thymic γδT and iNKT cell populations and found that Bcl11b^+/N797K progenitors were nearly outcompeted by control progenitors in generating these cells ( Figure 2C ). Previous studies have shown that Bcl11b plays an important role in presenting glycolipid antigens on CD1d during iNKT cell development (12). However, the distinct scarcity of mutant DP cells in our model rules out a role for impaired antigen presentation to iNKT-destined cells. As such, our results confirmed that a cell-autonomous mechanism was a major cause of impaired iNKT development in the Bcl11b^+/N797K mice.

We analyzed offspring obtained by intercrossing Bcl11b^+/N797K heterozygous mice and found that there were no live Bcl11b^N797K/N797K homozygous neonates at one week of age (P7), although Bcl11b^N797K/N797K homozygous mice were born alive at birth (P0) ( Supplementary Figure 3C ). This indicates that Bcl11b^N797K/N797 mice died during the neonatal period, as was observed in Bcl11b-defcient Bcl11b^Δ/Δ mice (27). Therefore, we examined thymocyte development in P0 neonates including Bcl11b^+/N797K obtained by crossing Bcl11b^+/N797K and Bcl11b^+/Δ mice, as well as Bcl11b^Δ/Δ mice. Compared to Bcl11b^+/+ and Bcl11b^+/Δ neonates, the total number of thymocytes was lower in Bcl11b^+/N797K neonates, and was further reduced in Bcl11b^Δ/N797K neonates ( Figure 3A ), in which only the mutant Bcl11b^N797K protein was expressed. Thymocytes cellularity was the lowest in Bcl11b^N797K/N797K and Bcl11b^Δ/Δ neonates and was comparable between these two genotypes ( Figure 3A ). The CD4 and CD8α expression profiles revealed a clear increase in the frequency of CD4⁻CD8⁻ DN thymocytes in the thymus of neonates harboring the Bcl11^N797K allele ( Figure 3A ). The generation of CD4⁺CD8⁺ DP thymocytes was significantly inhibited in Bcl11b^Δ/N797K neonates. A small CD4⁺CD8⁺ DP thymocyte subset was emerged in Bcl11b^N797K/N797K neonates, whereas these DP thymocytes were absent in Bcl11b^Δ/Δ neonates ( Figure 3A ). In line with the decrease in total thymocyte number, the number of CD4⁻CD8⁻ DN thymocytes was also decreased by the presence of the Bcl11b^N797K protein ( Figure 3B ). Moreover, CD44 and CD25 expression profile of the CD4⁻CD8⁻ DN population revealed that the frequency of CD44⁺CD25⁻ DN1 and CD44⁺CD25⁺ DN2 subsets was increased in Bcl11b^+/N797K , Bcl11b^Δ/N797K and Bcl11b^N797K/N797K neonates ( Figure 3B ). However, the cell number of these subsets was lower in all these mice than in wild-type mice. Thus, early thymocyte development in Bcl11b^N797K/N797K new-born mice was similar, but not identical, to that in Bcl11b^Δ/Δ mice, suggesting that Bcl11b^N797K variant protein retains its Bcl11b function to some extent.

To further examine T cell development at later stages from Bcl11b^N797K/N797 progenitors, we transplanted fetal liver (FL) cells prepared from Bcl11b^+/+ and Bcl11b^N797K/N797K at 13.5- or 14.5-dpc embryos into lethally irradiated Rag1-deficient recipients ( Supplementary Figure 3E ). Eight weeks after transplantation, we easily detected CD3ε⁺ cells in the peripheral blood of the recipients injected with Bcl11b^+/+ progenitors. In contrast, very few CD3ε⁺ cells were in the two recipients injected with Bcl11b^N797K/N797K progenitors, whereas their CD19⁺ B lymphocytes were well-reconstituted ( Supplementary Figure 3E ). The emergence of any CD3ε⁺ cells in the peripheral blood suggested that Bcl11b^N797K/N797K progenitors might retain some potency to give rise to mature T cells, prompting us to examine the lymphoid tissues. In the thymus of two recipients (#1 and #2), CD4⁺CD8⁺ DP thymocytes were reconstituted to some extent; however, as expected, the majority of CD4⁻CD8⁺ thymocytes were CD24⁺ ISP8 thymocytes ( Figure 4A ). Only a few mature CD24⁻TCRβ⁺ thymocytes and aberrant CD4⁻CD8α⁻ mature thymocytes emerged in these thymi ( Figure 4B ). In the CD4⁻CD8⁻ DN thymocyte population, the frequency of CD44⁺CD25⁺ and c-Kit⁺CD44⁺ cells was higher in recipients injected with Bcl11b^N797K/N797K progenitors than those injected with Bcl11b^+/+ progenitors ( Figure 4C ). In the c-Kit⁺CD44⁺ population, the frequency of CD25⁻ cells. which corresponds to ETP, was clearly higher in both recipients. Thus, engraftment occurred but was blocked in T-lineage development by the Bcl11b mutation.

In the spleen, the frequency of CD3ε⁺ T cells was approximately 1% in recipients injected with Bcl11b^N797K/N797K , while CD3ε⁺ T cells were reconstituted in more than 20% of Bcl11b^+/+ progenitors ( Figure 4D ). Among the CD3ε⁺ T cells generated from Bcl11b^N797K/N797K progenitors, γδTCR⁺ cells were almost undetectable. This finding indicates that the potency to generate T cells was not completely lost in Bcl11b^N797K/N797K progenitors. Interestingly, Bcl11b^N797K/N797K αβT cells showed dominant differentiation into CD4⁺ T cells with the emergence of CD4⁻CD8α⁻ αβT cells. Thus, Bcl11b^N797K/N797K progenitors failed to generate CD8⁺ T cells. Given that Bcl11b was shown to regulate the Zbtb7b gene, which encodes the CD4 helper-lineage specifying transcription factor ThPOK, we examined ThPOK expression via flow cytometry. We found that ThPOK expression was down-regulated not only in CD4⁻CD8⁻ cells but also in CD4⁺CD8⁻ αβT cells compared to Bcl11b^+/+ CD4⁺CD8⁻ αβT cells ( Figure 4E ). Therefore, the lack of CD8 SP mature cells was not because of overexpression of ThPOK.

A decline in CD45.1 frequency at the DN1 to DN2 transition in the competitive repopulation assay and an increase in the ETP frequency in Rag1-deficient recipients suggest that a developmental block occurs at the this transition in the presence of the Bcl11b^N797K mutant. To address this point further, we performed an in vitro T cell development assay using FL cells and OP9 stroma cells expressing a Notch ligand, delta-like 1 (OP9-DL1) ( Figure 5A ). Lineage-depleted 13.5-dpc FL cells were cultured on OP9-DL1 cells with supplement of cytokines, under conditions that support T-lineage development. Culture cellularities were calculated at all observation days, and frequencies of hemotopoetic cells were normalized to day 5 values to allow averaging across samples. The expansion of Bcl11b^N797K/N797K cells was slower than that of Bcl11b^+/+ cells ( Figure 5B ). CD4⁺CD8⁺ cells emerged from Bcl11b^+/+ cells on day 11, and the frequency of CD4⁺CD8⁺ cells increased by approximately 30% by day 14. On the contrary, no CD4⁺CD8⁺ cells emerged from Bcl11b^N797K/N797K cells, even at day 14, despite the appearance of a small but significant CD4⁻CD8⁺ subset. As a reference, we also cultured Bcl11b^Δ/Δ 17.5-dpc FL cells and found that the differentiation of CD4⁻CD8⁺ cells was almost undetected ( Figure 5C ). The higher potency of T cell development observed in Bcl11b^N797K/N797K cells than Bcl11b^Δ/Δ cells in culture is in line with the results obtained in recipient mice. Comparison of the frequency of DN1, DN2, DN3 and DN4 subsets during culture revealed a higher frequency of Bcl11b^N797K/N797K DN2 subset and a lower DN3 frequency from day 8 to day14 than Bcl11b^+/+ cells, indicating a developmental block at transition from DN2 to DN3 in Bcl11b^N797K/N797K cells. Interestingly, the frequency of the DN1 subset slightly increased from days 11 to 14 and was higher in Bcl11b^N797K/N797K cell cultures than Bcl11b^+/+ cells.

Considering the stepwise development of T cells from DN1 to DN2, DN2 to DN3 and DN3 to DN4, an increase in the CD44⁺CD25⁻ DN1 frequency in the later stage of Bcl11b^N797K/N797K culture was unusual but consistent with the higher frequency of CD44⁺CD25⁻c-Kit⁺ DN1-like cells observed in Rag1-defcient recipients. Given that Bcl11b^N797K/N797K DN2 cells showed a developmental arrest upon transition to DN3 cells, we examined the differentiation pathway of DN2 cells. To this aim, CD44⁺CD25⁺ DN2 cells were purified from Rag1-defcient recipients injected with either Bcl11b^+/+ or Bcl11b^N797K/N797K FL cells and cultured on TSt-4/Dll4 feeder cells. While Bcl11b^+/+ DN2 cells easily expanded and showed a 10-fold increase in their numbers within 5 days, most Bcl11b^N797K/N797K DN2 cells were dying, and less than 5% of the cells remained alive for 5 days in culture ( Figure 6A ). Flow cytometry analysis of living CD45⁺ cells showed the differentiation of CD4⁺CD8α⁺ DP cells from Bcl11b^+/+ but not Bcl11b^N797K/N797K DN2 cells. Interestingly, in the CD4⁻CD8α⁻ DN population, c-Kit⁺ cells emerged from Bcl11b^N797K/N797K cells but not Bcl11b^+/+ DN2 cells. Most c-Kit⁺ cells were negative for CD44 expression. CD44⁺ cells also emerged specifically from Bcl11b^N797K/N797K DN2 cells, and less than half of these CD44⁺ cells lost CD25 expression. Thus, c-Kit⁺CD44⁻ and c-Kit⁻CD44⁺ cells differentiated from Bcl11b^N797K/N797K DN2 cells at least in our culture condition.

Our results showed an increase in the frequency of ISP8 thymocytes by the Bcl11b^N797K variant protein. There are at least two mechanisms that lead to an increase in the ISP8 subset: a developmental block from the ISP8 to the CD4⁺CD8⁺ DP transition, or the attenuated activation of Cd4 gene, which causes CD4 downregulation in DP thymocytes. In order to characterize the ISP8 thymocytes that emerged in Bcl11b^+/N797K mice, we examined the expression of IL-7Rα, which was downregulated during the differentiation of CD4⁻CD8α⁻ DN into CD4⁺CD8α⁺ DP thymocytes with reduced chromatin accessibility at the Il7r promoter region, where Bcl11b binding was detected using ChIP-seq ( Supplementary Figure 4A, B ). Interestingly, both IL-7Rα expression level and chromatin accessibility at the Il7r promoter loci were lower in Bcl11b^+/N797K ISP8 cells than in Bcl11b^+/+ ISP8 cells ( Figures 7A, B ). In addition, Rorγt expression, which was induced from DN to ISP8 transition and slightly declined in DP thymocytes, was lower in Bcl11b^+/N797K ISP8 cells than in Bcl11b^+/+ ISP8 cells, and was similar between ISP8 and CD4⁺CD8⁺ DP thymocytes in Bcl11b^+/N797K mice ( Supplementary Figure 4C ). Thus, regarding Il7r and Rorc gene regulation, Bcl11b^+/N797K ISP8 cells showed characteristics similar to CD4⁺CD8α⁺ DP thymocytes, suggesting that impaired CD4 induction is involved in the accumulation of ISP8 cells in Bcl11b^+/N797K mice. Regarding Cd4 gene regulation, two transcriptional enhancers, proximal enhancer (E4p) (28, 29) and maturation enhancer (E4m) (24, 30), have been identified in the Cd4 locus. Genetic ablation of the E4p from the Cd4 locus showed that E4p is essential for inducing CD4 expression at the transition from CD4⁻CD8⁻ DN to CD4⁺CD8⁺ DP thymocytes (29). ATAC-seq results showed increased chromatin accessibility at both E4p and E4m regions during the DN to DP transition, where Bcll1b binding was also detected by ChIP-seq ( Supplementary Figure 4D ). Interestingly, chromatin accessibility in the E4m region was lower in Bcl11b^+/N797K ISP8 cells than in Bcl11b^+/+ ISP8 cells, while chromatin accessibility at the E4p region was comparable between these cells ( Figure 7C ).

The expression of the Cd4 gene was repressed by the intronic Cd4 silencer (S4) in both CD4⁻CD8⁻ DN and mature CD4⁻CD8⁺ SP thymocytes (31). The removal of the S4 from the Cd4 locus, referred to as the Cd4^+/ΔS allele (24), resulted in the de-repression of CD4 from CD25⁺ DN2/3 thymocytes and accelerated the CD4 upregulation at the transition from the DN to DP stage (32). Thus, CD4 induction at the DN-to-DP transition is regulated by both the activation of E4p and the inactivation of S4. To examine whether and to what extent the removal of the S4 reverses the increase in the ISP8 subset in Bcl11b^+/N797K mice, we generated Bcl11b^+/N797K:Cd4^+/ΔS double-mutant mice. While the removal of S4 reduced the frequency of CD4⁻CD8⁺ thymocytes in the CD24⁺TCRβ⁻ population to a nearly undetectable level in the Bcl11b^+/+ background, a small proportion of CD4⁻CD8⁺ thymocytes was still detected in Bcl11b^+/N797K: Cd4^+/ΔS mice. Thus, the genetic ablation of S4 did not completely reverse the increase in ISP8 cell frequency. This result suggests that the Bcl11b^N797K mutant inhibits Cd4 gene activation by affecting enhancer activity at the DN-DP transition stage or from an earlier stage. Given that Bcl11b gene is expressed at the DN2 stage and CD4 de-repression by loss of the S4 is induced in CD25⁺ DN cells, we examined CD4 expression levels in CD25⁺ DN cells and found that it was lower in Bcl11b^+/N797K: Cd4^+/ΔS mice than Bcl11b^+/+: Cd4^+/ΔS mice. Given that E4p is functionally responsible for CD4 induction in DP thymocytes, the Bcl11b^N797 mutant may interfere with the function of E4p in CD25⁺ DN cells. However, it is noteworthy that chromatin accessibility in DN thymocytes was reduced by the Bcl11b^N797K protein only at the E4m region, not at the E4p region ( Figure 7C ). This suggests that Bcl11b is involved in a step-wise activation of E4p enhancer after the chromatin structure is altered and that N797K replacement may attenuate such Bcl11b function.