The Ikzf3^N159S mutant mouse strain was generated by CRISPRCas9–mediated genome editing and has been reported (27). CD45.1 congenic mice and Rag1-deficient mice were obtained Jackson Laboratory. All mice were maintained in the animal facility at RIKEN Center for Integrative Medical Sciences, and all animal procedures were in accordance with institutional guidelines for animal care and the protocol approved by the Institutional Animal Care and Use Committee of RIKEN Yokohama Branch (2020–026).

Anti-Ikaros (MABE912) and anti-Aiolos (MABE911) purchased from Merck were used for western blot experiments. Antibodies and clone names used for flow cytometry were CD43 (S7), BP-1 (BP-1), IgM (R6-60.2), B220 (RA3-6B2), CD11b (M1/70), CD19 (1D3), CD24 (M1/69), CD93 (AA4.1), CD62L (MEL-14), CD44 (IM7), CD4 (RM4-5), CD8a (53-6.7), TCRβ (H57-597), CD3e (17A2), IgD (11-26c.2a), CD23 (B3B4), CD21 (7G6), CD5 (53-7.3), CD45.1 (A20), CD45.2 (104), Ly6G (1A8), Ly6C (HK1.4), CD11c (HL3), CD127 (A7R34) and NK1.1 (PK136) and were purchased from either BD Bioscience or BioLegend.

Flow cytometry experiments and fluorescence-activated cell sorting were performed on FACSCanto™ II (BD Biosciences) and FACSAria (BD Biosciences), respectively, and were analyzed by FlowJo software (Tree Star). Cell types were defined as follows: B cells (B220⁺CD19⁺), B-1 cells (CD19^hiB220^lo), B-2 cells (CD19^loB220^hi), B-1a cells (CD19^hiB220^loCD5⁺), B-1b cells (CD19^hiB220^loCD5⁻), T cells (TCRβ⁺CD3⁺), naïve T cells (CD62L⁺CD44⁻), central memory cells (CD62L⁺CD44⁺), effector memory cells (CD62L⁻CD44⁺), FO B cells (CD19⁺CD21^intCD23^hi), MZ B cells (CD19⁺CD21^hiCD23^lo/-). B cells in bone marrow, Fr.A (B220⁺CD43⁺Bp-1⁻CD24⁻), Fr.B (B220⁺CD43⁺Bp-1⁻CD24⁺), Fr.C (B220⁺CD43⁺Bp-1⁺CD24⁺), Fr.D (B220⁺CD43⁻AA4.1⁺IgM⁻), Fr.E&F (B220⁺CD43⁻AA4.1⁺IgM⁺) and recirculating B cells (B220⁺CD43⁻AA4.1⁻IgM⁺),dendritic cells (B220⁻TCRβ⁻CD11b⁺CD11c⁺), natural killer cell (B220⁻TCRβ⁻CD11b⁺CD11c⁻NK1.1⁺Ly6G⁻), neutrophils (B220⁻TCRβ⁻CD11b⁺CD11c⁻NK1.1⁻Ly6G⁺), eosinophils (B220⁻TCRβ⁻CD11b⁺CD11c⁻NK1.1⁻Ly6G⁻Ly6C^lo), monocytes/macrophages (B220⁻TCRβ⁻CD11b⁺CD11c⁻NK1.1⁻Ly6G⁻Ly6C⁻), inflammatory monocytes (B220⁻TCRβ⁻CD11b⁺CD11c⁻NK1.1⁻Ly6G⁻Ly6C^hi).

RNA-seq was performed using RNA extracted from purified splenic T cells (TCRβ⁺CD4⁺CD44⁻CD127⁺) or naïve B cells (B220⁺CD19⁺) from Ikzf3^+/+ (n=2 or 4), Ikzf3^+/N159S (n=2) and Ikzf3^N159S/N159S (n=2) mice. Total RNA was extracted by using Trizol reagent (Invitrogen). The mRNA was obtained by poly(A) selection with NEBNext Poly(A) mRNA Magnetic Isolation Module (E7490). RNA-seq libraries were prepared by NEBNext UltraIIRNA Library Prep Kit for Illumina (E7770) following the manufacturer’s protocol. Briefly, the first and second strands of cDNA were synthesized from mRNA, and were end-repaired, dA tailed and ligated with an adapter (E7335). The adapter-ligated DNA fragments were size selected by dual bead-based size selection using AMPure XP Beads (A63881). The size-selected DNA was amplified by PCR for 11–12 cycles. The purified DNA was diluted to 1 ng and was sequenced with a HiSeqX sequencer (Illumina) at Macrogen.

Trimmomatic is used to trim reads and removing adapter sequences. For analysis of differentially expressed genes (DEGs), RNA-seq reads were mapped to the murine genome (mm10) using HISAT2 (version 2.2.1) with default parameters, and read counts were calculated with HTseq (version2.0.2) using refGene as genome index files (28). Generated read counts were analyzed with the R package edgeR (version 3.38.4) (29). For gene ontology (GO) analysis, downregulated DEGs with false discovery rate (FDR) <0.05 in Ikzf3^N159S/N159S T cells or B cells compared to Ikzf3^+/+ controls were extracted and subjected to the analysis. GO enrichment analysis was performed with the R package clusterProfiler (version 4.4.4) (30).

And the ENCODE database showed the Ikaros protein binds to SELL gene in human B-Lymphocyte cell line (GM12878 cell line). And Harmonizome databases extract and organize the genes or proteins from publicly available resources (31). In Harmonizome databases, we search and confirm which down-regulated genes of RNA-seq data are the target genes of the IKZF1 transcription factor in ChIP-seq datasets, and then verified them in the ENCODE database.

We constructed retroviral vectors for generating retroviruses to infect mouse B cells. The entire coding region of mouse Ikaros/Ikzf1 (NM_001025597.3) was amplified from cDNA prepared from wild-type splenic cells by RT-PCR and was cloned into pCR2.1^®-TOPO vector by TOPO TA Cloning Kit (Invitrogen). After confirming the sequence, the cDNA fragment was cloned into the pMigR-DsRED plasmid, which was generated by ourselves via replacing the GFP sequence with DsRED in the pMigR1 plasmid (Addgene Plasmid #27490). Similarly, cDNA encoding mouse Aiolos/Ikzf3 (NM_011771.1) was amplified by RT-PCR and was cloned into pMigR1 plasmid. Primers sequences were following: Ikzf1-Forward : CCCAGATCTAGACAATGGATGTCG-ATGAGG, Ikzf1-Reverse : CCCCTCGAGGGTTTAGCTCAGGTGGTAAC, Ikzf3-Forward : CCCGCGGCCGCCGGCGACATGGAAGATATAC, Ikzf3-Reverse : CCCCTC GAGTGCTCACTTCAACATGGCTC

Plate-E packaging cell line was cultured in DMEM medium (Gibco) supplemented with 10% FBS (Hyclone), 1% penicillin/streptomycin (Gibco), 10 µg/ml Blasticidin (Gibco) and 1 µg/ml Puromycin (Gibco) at 37°C with 5% CO₂. Plate-E cells (2*10⁵/ml in a 6-well dish) were transfected with 2 µg of the retroviral vector using a FuGENER HD Transfection Reagent (Promega) at a 3:1 ratio (FuGENER HD: DNA mass per well). After 48h, the supernatant was collected. B cells from WT and Ikzf3^N159S/N159S mice were isolated using B cell Microbead (Miltenyi Biotec) with purity over 92%. 10⁶ of these B cells were stimulated with 10 µg/ml LPS (Sigma) in RPMI medium (Gibco), supplemented with 10% FBS, 50 µmol 2-mercaptoethanol (Sigma), 1mM sodium pyruvate (Gibco), 1% penicillin/streptomycin (Gibco) and 1% MEM (Gibco). Two days after stimulation, viral supernatant was added with 20 mg/ml polybrene (Greiner bio-one) to the cells in a 24-well plate. To enhance transduction efficiency, cells were subjected to a round of centrifugation with vector-containing medium for 90min at 2000rpm and 32°C, separated by a 2-hour interval in which cells were cultured in medium without viral particles. Two days after infection, the cells were FACS analyzed for various purposes as indicated.

For the generation of bone marrow chimeras, bone marrow (BM) cells were prepared from CD45.2 WT or Ikzf3^N159S/N159S mice and CD45.1 wild type mice. Those donor-derived CD45.2 bone marrow cells were mixed 1:1 with bone marrow cells from CD45.1⁺ wild type cells and were transferred intravenously into sub-lethally (950 rad) irradiated Rag1^−/− recipients. The recipient mice were analyzed 8–12 weeks after bone marrow injection.

The homing assay was performed as described (32–35) with slight modifications. Single splenic cell suspensions were prepared from CD45.2 WT or Ikzf3^N159S/N159S mice and CD45.1 WT mice. Equal numbers (3-5 × 10⁷) of CD45.2 WT or Ikzf3^N159S/N159S cells and CD45.1 WT cells were mixed and were intravenously co-injected into CD45.1/2 recipient mice. An aliquot of the injected cells mixture was analyzed by flow cytometry to confirm the injected ratio of CD45.1⁺ and CD45.2⁺ (Ri) in B220⁺CD19⁺ or TCRβ⁺CD3⁺ population. After 24h of migration, single-cell suspensions of blood and organs were prepared and were stained with antibodies. The ratio of CD45.1⁺CD45.2⁻ and CD45.2⁻CD45.2⁺ (Ro) was determined by flow cytometry. The ratio of CD45.1⁺CD45.2⁻ and CD45.2⁻CD45.2⁺ cells in B220⁺CD19⁺ or TCRβ⁺CD3⁺ population within individual injection mixture (Ri) and organs (Ro) was measured, and the homing assay results were presented as the ratio of Ro/Ri in each tissue.

Given the absence of experimentally determined three-dimensional structure for human AIOLOS, a homology model of the four N-terminal zinc fingers was constructed. Initially, a template search was performed using BLASTP against the RCSB protein data bank (PDB) with the human AIOLOS sequence as the query, leading to the identification of potential template sequences. After careful evaluation, a template fragment of human PR/SET domain 9 (PRDM9c) with a PDB ID of 5V3G, encompassing six zinc fingers, was selected (36–38). The choice was based on its sequence identity of 38%, absence of gaps (0% gap), 55% positive residues, and a high BLAST bit score of 224.

To generate the homology models, MODELLERv9.17 was employed (39). Initially, the target and template sequences were aligned using MODELLER’s align2d command, which employs global dynamic programming with a linear gap penalty function to align the two profiles. Subsequently, 20 three-dimensional models of human AIOLOS with four N-terminal zinc fingers and DNA were built using MODELLER. Among the generated models, the one with the lowest DOPE (Discrete Optimized Protein Energy) score, as well as visual inspection, was chosen for further analysis (40). This selected model represents the complex of human AIOLOS with four N-terminal zinc fingers, cognate DNA sequence of AIOLOS, and zinc ions, and served as the starting structure for subsequent molecular dynamics simulations and analyses.

To investigate the behavior of the N160S as compared to wild-type (WT) and G159R-mutant structures, we first prepared the N160S-mutant by mutating the N160 to S160 in silico, while preserving the secondary structure. Subsequently, molecular dynamics (MD) simulations were performed. The initial step involved the addition of hydrogen atoms to all the systems. Subsequently, these structures were solvated using the three-point transferable intermolecular potential (TIP3P) water model in an octahedral box, while maintaining a distance of approximately 10 Å between the protein-DNA complex surface and the box boundary (41). The systems were then neutralized by introducing counter ions, and the simulations were performed using GROMACS 5.1.2 with the ff14SB forcefield (42, 43). The MD simulations commenced with an energy minimization phase, consisting of 50,000 steps, followed by a gradual heating process from 0 to 300 K over 500 ps. Subsequently, a constant temperature equilibration was performed for 3000 ps at 300 K. To ensure velocity stability, the Parrinello-Rahman barostat pressure coupling method was employed. The production runs for the WT, N160S and G159R mutants lasted 500 ns, employing a periodic boundary condition in the NPT ensemble. Berendsen temperature coupling with modifications was utilized to maintain a constant temperature of 300 K and a pressure of 1 atm. The leap-frog integrator and the Verlet cut-off schemes were employed during the simulations (44). To preserve bond lengths, the LINCS algorithm was used, while the Particle-mesh Ewald method was employed for the calculation of electrostatic forces. In these calculations, Fourier grid spacing and Coulomb radius values of 0.16 nm and 1.4 nm, respectively, were utilized. The standard van der Waals interactions were set to 1.4 nm. To analyze the MD simulated trajectories of the three systems, several key metrics were calculated. The root means square deviation (RMSD), root mean square fluctuation (RMSF), and radius of gyration (Rg) were determined, for which the GROMACS utilities, namely gmx rms, gmx rmsf, and gmx gyrate, were employed for these calculations respectively, providing essential insights into the structural stability and flexibility of the systems. The MD simulation trajectories were visualized using Visual Molecular Dynamics (VMD) (45). Molecular visualization and figure generation were accomplished using PyMOL (http://www.pymol.org).

We have previously reported that the knock-in mouse model carrying the Ikzf3^N159S mutant allele generating the Aiolos^N159S missense mutant protein, which corresponds to the human AIOLOS^N160S variant, recapitulated the abnormal B and T lymphocyte development similarly to what was observed in the patients with IKZF3^N160S variant. B cell frequencies in peripheral blood (PBL) was decreased in Aiolos^N159S/N159S mice (27). In addition, we showed that the expressions of IgM, IgD, and CD23 significantly decreased in B lymphocyte populations in both Ikzf3^+/N159S and Ikzf3^N159S/N159S mice (27). To clarify the developmental stages of B cell development impaired by the Aiolos^N159S mutation, we examined early B cell development in the bone marrow according to the criteria of Fractions A to G defined by Hardy (46). No significant differences were observed in the frequencies of fractions A-C among Ikzf3^+/+ , Ikzf3^+/N159S and Ikzf3^N159S/N159S mice ( Figures 1A, B ). However, in the CD11b⁻B220⁺CD43⁻ population, AA4.1⁺IgM⁺ cells corresponded to Fr. E & F were increased in Ikzf3^N159S/N159S mice with a decrease in recirculating B cells, which were defined as CD11b⁻B220⁺CD43⁻AA4.1⁻IgM⁺ ( Figures 1A, B ). In contrast, recirculating B cells increased in Ikzf3^+/N159S mice ( Figures 1A, B ), which is consistent with the tendency of increased splenic B cell frequency in Ikzf3^+/N159S mice (27). Therefore, an increase in Fr. E & F in Ikzf3^N159S/N159S mice might reflect a relative increase due to the low frequency of recirculating B cells, rather than the accumulation of immature B cells by developmental blockade.

B-lymphocytes are divided into B1 and B2 subtypes. B1 cells are an important part of the innate immune response, and spontaneously secrete polyreactive IgM antibodies (47, 48). In mice, B1 cells differ from conventional B2 B cells in that they do not express CD21 and CD23 and express only low levels of B220 (B220^loCD19^hi) and IgD, but high levels of IgM. B1 cells are abundant in both peritoneal and pleural cavities and can be further divided into B1a (B220^loCD19^hiCD5⁺) and B1b (B220^loCD19^hiCD5⁻) cells (47, 48). CD5⁺ B1a cells are thought to be long-lived, self-renewing cells that predominantly reside in these pleural and peritoneal cavities, where they produce natural polyreactive IgM antibodies with a biased, autoreactive repertoire (49).

Given the reduction in B2 cell numbers in Ikzf3^N159S/N159S mice, we examined whether B1 cell development was impaired by the Ikzf3^N159S mutation. Consistent with the tendency of splenic B2 cell numbers (27), the frequency of B2 cells, defined as CD19^loB220^hi cells in the peritoneal cavity of Ikzf3^+/N159S and Ikzf3^N159S/N159S mice, increased and decreased, respectively, thereby increasing and decreasing the B2 to B1 ratio ( Figures 1C, D ). B2 cells in the peritoneal cavity of both mutants also showed decreased IgD and CD23 expression ( Figure S1A ), as observed in the spleen (27). In the CD19^hiB220^lo B1 cell population, the expression level of IgD was also decreased by the Ikzf3^N159S mutation in a dose-dependent manner, and IgD⁺ cells were hardly detected in Ikzf3^N159S/N159S mice ( Figures 1C, D ). B1a cells, defined as CD19^hiB220^loCD5⁺ cells, were significantly decreased in the peritoneal cavities of Ikzf3^+/N159S and Ikzf3^N159S/N159S mice ( Figures 1C, D ). Since CD5 is the only marker to distinguish B1a and B1b cells, loss of B1a cell population could be either caused by loss of CD5 expression from B1a cells or developmental block of B1a cells. In order to distinguish whether CD5 expression was lost in B1a cells or whether B1a cell development was inhibited by Aiolos^N159S, we examined CD5 expression in other cells such as T lymphocytes. In the spleen, the frequency of CD5⁺ T cells were decreased in Ikzf3^+/N159S mice with severer reduction in CD8⁺ T cells ( Figure S1B ). Primary T cell development in the thymus is divided into four stages according to CD4 and CD8 co-receptor expression: most immature CD4⁻CD8⁻ (DN) stage is flowed by CD4⁺CD8⁺ (DP) precursor stage, and DP precursor is maturated into CD4⁺CD8⁻ (CD4 SP) or CD4⁻CD8⁺ (CD8 SP) stage after positive selection. In thymic T cells, the frequency of CD5 expressing cells was significantly decreased in the CD4⁺CD8⁺ (DP) and CD4⁻CD8⁺ (CD8 SP) subsets and increased in the CD4⁻CD8⁻ (DN) subset ( Figure S1C ). This indicated that Ikzf3^N159S mutation disturbed CD5 expression during the DN to DP stages in the thymus, specifically in the CD8 SP population. However, CD5 expression was not completely lost, suggesting that the loss of CD5⁺ B1a population was likely caused by developmental inhibition rather than a loss of CD5 expression.

Although just from a single patient, RNA-seq analyses of naïve B cell from such a patient carrying IKZF3^N160S variant showed lower expression of SELL gene which encodes CD62L (27). In Ikzf3^N159S/N159S mice, circulating T cells in the peripheral blood, particularly CD8⁺ T cells, showed decreased CD62L expression (27). We then next examined the developmental stages that initiate low CD62L expression during T cell development. In primary T cell development of thymus, flow cytometry analyses detected low CD62L expression of thymocytes of Ikzf3^+/N159S and Ikzf3^N159S/N159S mice from CD4⁺CD8⁺ (DP) stage onward, while CD62L expression on CD4⁻CD8⁻ DN thymocytes was comparable ( Figures 2A, B ). In mice, RNA-seq data available in the ImmGen database showed that the expression of Ikzf3 gene is initiated from the DN to DP transition, whereas the Ikzf1 gene is expressed earlier in the DN1 stage ( Figure S2A ). Thus, low CD62L and CD5 expression was correlated with the expression of Ikzf3^N159S gene, suggesting that the Aiolos^N159S mutant protein quickly suppresses CD62L expression in a dominant-negative form. Given that CD62L is expressed in mature B cells, we also examined CD62L expression in splenic B220⁺ CD19⁺ B cells and found that expression of CD62L is heavily decreased in Ikzf3^+/N159S and Ikzf3^N159S/N159S mice ( Figures 2C, D ). Furthermore, B220⁺CD19⁺ cells in the peripheral blood and peripheral LN (pLN), and mesenteric LN (mLN), and B2 and B1 cells in the peritoneal cavity showed decreased CD62L expression in Ikzf3^N159S/N159S mice ( Figures S2B, C ). Interestingly, the frequency of CD62L⁺ B cells varied in secondary lymphoid tissues of Ikzf3^+/N159S mice.

We observed that the differentiation of T and B lymphocytes, NK1.1⁺ natural killer (NK) cells, and eosinophils (Eos) was impaired with the loss of certain markers such as CD23 and CD62L in Ikzf3^N159S/N159S mice (27). To address whether these B, T, and myeloid cell phenotypes occur in a cell-intrinsic manner and to test the developmental potency of progenitors under competitive settings, we performed mixed bone marrow (BM) chimera experiments. CD45.1⁺ bone marrow cells were mixed with CD45.2⁺ bone marrow cells prepared from either Ikzf3^+/+ or Ikzf3^N159S/N159S mice in equal proportions and transplanted into sublethally irradiated Rag1 KO recipient mice ( Figure S3A ). Reconstitution of hematopoiesis in recipients was analyzed 8 weeks after transplantation using flow cytometry. In the total bone marrow cells, the CD45.2⁺/CD45.1⁺ ratio was close to 1:1 in recipients injected with CD45.1⁺ Ikzf3^+/+ and CD45.2⁺ Ikzf3^N159S/N159S BM cells ( Figures 3A, B ). However, during early B cell development, the percentage of CD45.2⁺ Ikzf3^N159S/N159S cells increased in Fr. E and F ( Figures S3B, C ), suggesting that mature B cells in the bone marrow were retained or accumulated in the presence of Aiolos^N159S. In B220⁺CD19⁺ B cell population in peripheral lymphoid tissues, the percentage of CD45.2⁺ Ikzf3^N159S/N159S cells was significantly decreased to approximately 10% in spleen and was further reduced to less than 1% in pLN, mLN, and peritoneal cavity, comparing with CD45.2⁺ Ikzf3^+/+ cells ( Figures 3A, B and S3D ).

In the TCRβ⁺ T cell population, CD45.2⁺ Ikzf3^N159S/N159S cells were significantly decreased in the peripheral and mesenteric lymph nodes compared to CD45.2⁺ Ikzf3^+/+ cells ( Figures 3C, D and S3E ). In NK and myeloid lineage cells, the frequency of CD45.2⁺ Ikzf3^N159S/N159S cells was lower, specifically in CD11b⁺NK1.1⁺ NK cells and CD11b⁺Ly6G⁻Ly6C^lo Eos cells. ( Figures 3E, F and S3F, G ). These observations indicate that developmental defects of NK and Eos by the Ikzf3^N159S mutation were caused in a cell-intrinsic manner and developmental potency to other myeloid-lineage cells, at least cell types we examined, were not affected by the Ikzf3^N159S mutation.

Next, we examined whether the impaired expression of surface proteins caused by the Ikzf3^N159S mutation occurs in a cell-intrinsic or cell-extrinsic manner. In recipients transplanted with CD45.1⁺ Ikzf3^+/+ and CD45.2⁺ Ikzf3^N159S/N159S BM cells, the expressions of IgD, CD23, and CD21 in the splenic B cell population were heavily reduced, specifically in CD45.2⁺ Ikzf3^N159S/N159S cells ( Figures 4A, B ). Similarly, CD62L expression was significantly decreased, specifically in CD45.2⁺ Ikzf3^N159S/N159S thymocytes ( Figures 4C, D ). These results indicated that the Aiolos^N159S mutant protein inhibits the expression of IgD, CD23, CD21, and CD62L in a cell-intrinsic manner. Interestingly, the frequency of CD62L⁺ cells in each thymocyte subset was higher in these recipients than in the non-irradiated control mice ( Figures 2A, B and 4C, D ). This suggests that the reconstitution of T cell development in the thymic environment in irradiated recipients may induce CD62L. It is noteworthy that the frequency of CD62L⁺ in CD4⁻CD8⁻ DN subset was already lower in CD45.2⁺ Ikzf3^N159S/N159S thymocytes compared with CD45.2⁺ Ikzf3^+/+ , suggesting that a small amount of Aiolos^N159S may shortly inhibit CD62L expression after induction in DN cells ( Figures 4C, D ). As expected, CD62L expression was severely inhibited in T cells from the DP thymocytes stage onward and in the spleen ( Figures 4E, F ). However, in pLNs, although the CD45.2⁺ Ikzf3^N159S/N159S T cells comprised a small population, the frequency of CD62L⁺ cells was comparable to that of control cells, similar to CD45.2⁺ Ikzf3^N159S/N159S CD4 T cells in mLN ( Figures 4E, F and S4E, F ).

Our mixed bone marrow chimera experiments clearly showed a greater reduction in the frequency of CD45.2⁺ Ikzf3^N159S/N159S lymphocytes in the LNs than in the spleen, whereas the frequency of CD62L⁺ T cells increased in the LNs. Given the important role of CD62L in lymphocyte homing to lymphoid tissue, we compared the homing abilities of Ikzf3^+/+ and Ikzf3^N159S/N159S lymphocytes. Total splenocytes from CD45.2⁺ Ikzf3^+/+ or CD45.2⁺ Ikzf3^N159S/N159S mice and CD45.1⁺ control mice were mixed in a 1 to 1 ratio and intravenously injected into CD45.1⁺CD45.2⁺ recipient mice ( Figure 5A ). After 24 hours, the frequencies of CD45.1⁺ and CD45.2⁺ cells in the PBL, spleen, and LNs were analyzed by flow cytometry. The ratio of CD45.1⁺CD45.2⁻ and CD45.2⁻CD45.2⁺ in B220⁺CD19⁺ or TCRβ⁺CD3⁺ population in recipient organs (Ro) and that in injection mixture (Ri) were used to calculate the homing index as the ratio of Ro/Ri ( Figure 5A ). In recipients injected with CD45.1⁺ control cells and CD45.2⁺ Ikzf3^+/+ , the homing index was approximately one in all tissues examined ( Figures 5B–E and S5A ). In contrast, in recipients injected with CD45.1⁺ control cells and CD45.2⁺ Ikzf3^N159S/N159S cells, the homing index in the B cell compartment was lower than 0.2 in the PBL, although it remained around 0.5 in spleen, and CD45.2⁺ Ikzf3^N159S/N159S B cells were hardly detected in the LNs ( Figures 5B, C and S5A ). In the T cell population of the above recipients, the homing index was decreased in the pLN and mLN, but not in the PBL and spleen ( Figures 5D, E and S5A ). These observations indicate that the homing ability to the LNs was impaired in Ikzf3^N159S/N159S B and T lymphocytes. As expected from analyses of mixed bone marrow chimeras, the frequency of CD62L⁺ cells in Ikzf3^N159S/N159S B and T cell populations was significantly increased in pLN and mLN from the input, although these were still lower than those in CD45.2⁺ Ikzf3^+/+ cells ( Figures 5B–E and S5A ). These results revealed that Ikzf3^N159S/N159S B and T lymphocytes were outcompeted by control cells in homing to the LNs. Selective accumulation of CD62L⁺ Ikzf3^N159S/N159S lymphocytes in LNs revealed that Ikzf3^N159S/N159S lymphocytes expressing CD62L retained their homing activity to some extent, suggesting that the Aiolos^N159S mutation impaired the homing ability of lymphocytes, mainly via disturbing the CD62L expression.

To examine what genes were dysregulated in the presence of Aiolos^N159S variant, we conducted RNA-seq analyses. Considering low CD62L expression in T cells, we chose CD127 as a marker to isolate cells that would correspond to naïve T cells and purified splenic B220⁺CD19⁺ B cells and TCRβ⁺CD4⁺CD44⁻CD127⁺ T cells of Ikzf3^+/+ , Ikzf3^+/N159S , and Ikzf3^N159S/N159S mice. The results obtained from RNA-sequencing (RNA-seq) were analyzed using principal component analysis (PCA). In both B and T cells, PCA showed that Ikzf3^+/+ , Ikzf3^+/N159S , and Ikzf3^N159S/N159S cells were differentially clustered in terms of global gene expression ( Figure 6A ). Gene ontology (GO) analysis was subsequently performed using downregulated genes in Ikzf3^N159S/N159S B and T cells. GO analysis indicated that the downregulated genes were enriched in lymphocyte activation, differentiation, and proliferation, leukocyte migration, and cell adhesion ( Figure 6B ). In the heat map, we defined the downregulated genes of leukocyte migration in Ikzf3^N159S/N159S B cells, showing an intermediate phenotype in Ikzf3^+/N159S B cells between WT and homo B cells in terms of gene expression ( Figure 6C ). In Ikzf3^+/N159S T cells, most of genes in the leukocyte migration term also showed an intermediate phenotype, except for CD24a and Cnr2 genes ( Figure 6D ). This observation provides evidence of the gene dosage effect of the Aiolos^N159S variant on transcriptional dysregulation. We also examined our RNA-seq data for expression of genes encoding chemokine receptors and found that some chemokine receptors Ccr6, Ccr7, Ackr2 and Ccr1 were downregulated in Ikzf3^N159S/N159S B cells, while only Cxcr1 showed a minimal increased expression of Ikzf3^N159S/N159S T cells ( Figure S6 ).

While CD62L expression in CD4⁺ T cells was not reduced by Aiolos deficiency (50), loss of Ikaros function led to reduced CD62L expression in naïve T cells due to low Foxo1 expression (51). However, Foxo1 expression was not reduced in B or T cells by the Ikzf3^N159S mutation ( Figure 6E ). We have shown that another murine Aiolos variant, Aiolos^G158R, which has a G to R replacement next to the N159 residue, interferes with Ikaros function (16). To explore the possibility that the genes downregulated by Aiolos^N159S were correlated with Ikaros function, we utilized the Harmonizome databases, which is a collection of processed datasets gathered to serve and mine knowledge about genes and proteins from over 70 major online resources (31). Although the Harmonizome database does not have sufficient datasets to predict IKZF3/AIOLOS target genes, IKZF1/IKAROS target genes can be predicted using chromatin immunoprecipitation sequencing (ChIP-seq) datasets of IKZF1 from the ENCODE Transcription Factor Targets datasets. Our analyses of 16 and 31 dysregulated genes in T and B cells, respectively, found that four genes in T cells and more than half of the genes in B cells were associated with IKAROS protein binding ( Figure 6F ). The genome browser track of the ENCODE IKAROS ChIP-seq datasets in the human B cell line showed potential binding of IKAROS to the SELL locus encoding CD62L ( Figure 6G ). Thus, transcriptome analyses suggested that the downregulation of some genes by the Aiolos^N159S variant was caused by impaired Ikaros function.

Given the gene dosage-dependent effect of Ikzf3^N159S on CD62L expression, when low CD62L expression was caused by the interference of Ikaros by Aiolos^N159S, an increase in the ratio of Ikaros to Aiolos^N159S may overcome the dominant effect of the Aiolos^N159S variant. To test this possibility, we constructed a retroviral vector encoding either Ikaros with a DsRed marker or Aiolos with a GFP marker. After confirming the presence of Ikaros and Aiolos proteins in the packaging cell line Plat-E by western blotting (WB) ( Figure 7A ), we transduced splenic Ikzf3^+/+ and Ikzf3^N159S/N159S B cells by retroviral transduction. Induction with Ikaros significantly increased the frequency of the CD62L⁺ population in Ikzf3^+/+ cells. However, Aiolos transduction decreased CD62L expression compared to the empty vector (EV) ( Figures 7B, C ). Importantly, CD62L expression in Ikaros transduced (DsRED⁺) Ikzf3^N159S/N159S B cells was partially restored to approximately 20% ( Figures 7B, C ), although the CD62L expression remained lost after Aiolos transduction. These findings support the hypothesis that Aiolos^N159S interferes with Ikaros function, at least in the transcriptional regulating of CD62L expression.