This study is compliant with all the relevant ethical regulations. All mouse experiments were performed under protocols approved by National Institute on Aging (NIA) Institutional Animal Care and Use Committees (338-LMBI-2022).

Mice

Young (8–12 weeks) recombination activating gene 2-deficient (Rag2−/−) mice35,36 were purchased from Jackson Lab (stock no. 008449) and maintained to be old (100–110 weeks) in house by NIA Comparative Medicine Section. Young (8–12 weeks) and old (100–110 weeks) C57BL/6J mice were provided by the NIA Aged Rodent Colonies. The mice used in the experiments comprised both males and females, and for each experiment, the sexes of both young and old mice were matched. In each experiment, two to six mice were pooled on the basis of the required cell count and the actual number of cells obtained.

Mouse husbandry was conducted under standard circadian rhythms of 12 h of darkness followed by 12 h of light, at a constant temperature of 22 ± 2 °C and humidity ranging between 45% and 70%.

Cell lines

Rag2−/−Ebf1+/+Pax5+/+, Rag2−/−Ebf1+/−Pax5+/+ and Rag2−/−Ebf1+/−Pax5+/− pro-B cells are cultured as previous described40.

The wild-type D345 cell line was generously provided by David G. Schatz of Yale University. These cells were maintained in RPMI1640 (Gibco) supplemented with 10% foetal bovine serum (FBS, Gemini) and a 1× penicillin–streptomycin solution (Gibco).

293T human embryonic kidney cells used for lentiviral expression was purchased from ATCC (cat. no. CRL-3216) and cultured in Dulbecco’s modified Eagle medium (Gibco) with 10% FBS (Gemini) and 1× penicillin–streptomycin solution (Gibco).

The Platinum-E (Plat-E) cell line used for retroviral expression was purchased from Cell Biolabs (cat. no. RV-101) and cultured in Dulbecco’s modified Eagle medium (Gibco) with 10% FBS (Gemini), 1× penicillin–streptomycin solution (Gibco), 10 mM HEPES and 0.3 μg ml−1l-glutamine. All cells were cultured at 37 °C in a humidified atmosphere with 5% CO2.

AntibodiesFlow cytometry

Phycoerythrin (PE) anti-CD19 (BioLegend, 6D5, cat. no. 115508, 1:100), Brilliant Violet 421 (BV421) anti-B220 (BioLegend, RA3-6B2, cat. no. 103240, 1:100), fluorescein isothiocyanate (FITC) anti-IgM (BioLegend, RMM-1, cat. no. 406506; 1:50), PE anti-CD43 (BD Biosciences, S7, cat. no. 553271, 1:100), H3K27ac-AF488 conjugated (Cell Signaling Technology, cat. no. 15485S. 1:50) and IgG isotype control-AF488 conjugated (Cell Signaling Technology, cat. no. 4340S; 1:50) were used.

ChIP–seq and HiChIP

Anti-H3K27ac (Active Motif, cat. no. 39133; 1:500), anti-H3K27me3 (Diagonode, cat. no. C15410195; 1:500), anti-CTCF (Abcam, cat. no. ab70303; 1:250), anti-Rad21 (Abcam, cat. no. ab992; 1:500), anti-Brg1 (Abcam, cat. no. ab110641; 1:200) and anti-p300 (Cell Signaling Technology, cat. no. 57625S; 1:333) were used.

Pro-B cell purification

For Rag2−/− mice, total bone marrow was extracted from tibia and femurs, and erythrocytes were lysed. Pro-B cells were purified by combining positive selection using CD19+ selective beads (Stem Cell Technology, cat. no. 18954) and sorting by CD19+ and B220+ markers.

For C57BL/6J mice, total bone marrow was extracted from tibia and femurs, and erythrocytes were lysed. Cells were pre-purified using CD19+ selective beads (Stem Cell Technology, cat. no. 18954) and sorted with IgM−B220+CD43+ markers. Flow cytometry experiments were performed on BD FACSAria II Cell Sorter. FlowJo software was used for data analysis.

In situ Hi-C

Genome-wide in situ Hi-C was performed with Rag2−/− young and old primary pro-B cells, and Rag2−/−Ebf1+/+Pax5+/+, Rag2−/−Ebf1+/−Pax5+/+, Rag2−/−Ebf1+/−Pax5+/−, D345-BFP control and D345-Wapl-BFP pro-B cell lines, using the Arima Hi-C Kit (Arima Genomics), including KAPA Hyper Prep indexing and library amplification (cat. no. KK8500, Roche Molecular Systems Inc) according to the manufacturer’s instructions. For each assay, 1 × 106 cells were used as the input materials and two biological replicates were performed for each group. Samples were sequenced 2 × 150 bp on an Illumina NovaSeq instrument at the Single Cell and Transcriptomics Core at Johns Hopkins University.

RNA-seq

Total RNA was prepared using the Direct-zol RNA Miniprep Kit (Zymo Research, cat. no. R2051) according to manufacturer’s instructions. The RNA libraries for young and old Rag2−/− primary pro-B cells (four biological replicates for each group) were prepared using the NEXTFLEX Rapid Directional RNA-Seq Kit (PerkinElmer, cat. no. NOVA-5138-07) and sequenced 2 × 75 bp on an Illumina NovaSeq instrument at the Single Cell and Transcriptomics Core at Johns Hopkins University.

RNA libraries for Rag2−/−Ebf1+/+Pax5+/+, Rag2−/−Ebf1+/−Pax5+/+, Rag2−/−Ebf1+/−Pax5+/− and D345-BFP control and D345-Wapl-BFP pro-B cell lines (two replicates for each group) were prepared using SMARTer Stranded Total RNASeq Kit v2 (Takara, cat. no. 634412) and sequenced 1 × 100 bp on an Illumina NovaSeq instrument at the Single Cell and Transcriptomics Core at Johns Hopkins University.

RNA libraries for pro-B cells infected with Ebf1-GFP or GFP-containing retrovirus were prepared using the SMART-Seq mRNA kit (Takara, cat. no. 634773) and sequenced 2 × 100 bp on an Illumina NovaSeq instrument at the Computational Biology and Genomics Core in NIA.

ChIP–seq

For the ChIP of histone modifications, CTCF and Rad21, both young and old primary Rag2−/− pro-B cells underwent crosslinking with 1% formaldehyde (Sigma) for 10 min at room temperature. The reaction was subsequently quenched with 125 mM glycine, and cell lysis was initiated in a buffer containing 1% sodium dodecyl sulfate. Chromatin was then sheared using a Bioruptor (Diagenode) in cycles of 30 s on and 30 s off, totalling 15 min of shearing time. Immunoprecipitation was performed using specific antibodies, with 1 × 106 cells used for each assay, and antibody dilutions followed established protocols.

As for Brg1 and p300 ChIPs, we followed the protocol by Bossen et al.66, with some modifications. In this case, two million young and old primary Rag2−/− pro-B cells were first crosslinked with 1.5 mM ethylene glycol bis(succinimidyl succinate) at room temperature with rotation for 15 min. Subsequently, the cells were fixed with 1% formaldehyde (Sigma) for 15 min at room temperature with rotation and quenched with 200 mM glycine. After two phosphate-buffered saline (PBS) washes, the cells were lysed in a buffer containing 1% sodium dodecyl sulfate. Notably, we extended the sonication time to a total of 20 min of shearing time, while the subsequent steps of the protocol mirrored the previously mentioned procedure.

RT–PCR

Total RNA was isolated as previously described. A total of 100 ng of RNA was used to generate complementary DNA (cDNA) with SuperScript III (Thermo Fisher Scientific, cat. no. 18080051) using random hexamers according to the manufacturer’s protocol. Quantitative polymerase chain reaction (RT–PCR) was performed with iTaq Universal SYBR (Bio-Rad, cat. no. 1725125) using primers described in Supplementary Table 6. Three independent experiments were carried out. Cobp2 was used as a control37. Data were processed by the delta–delta CT method.

FISH

FISH was performed as previously described45,46. Briefly, cells were fixed on poly-l-lysine coated slides using 1.5 × 106 cells per slide. Slides were kept for 15 min at 37 °C, followed by fixation with 4% paraformaldehyde in PBS. Cells were further washed with 0.1 M Tris–HCl (pH 7.4), followed by PBS. Fixed cells were treated with 100 μg ml−1 RNase A in PBS, permeabilized using 0.5% saponin/0.5% Triton X-100/PBS for 30 min at room temperature. Probes were listed in Supplementary Table 6.

Labelled probes were denatured at 75 °C for 5 min and applied to cells that were denatured in formamide at 73 °C, followed by incubation in a dark humid chamber. Images were acquired using a Nikon Eclipse Ti2 microscope equipped with a 60× lens. Depending on the size of the nucleus, 40–50 serial optical sections spaced by 0.2 μm were acquired. The datasets were deconvolved using NIS-Elements software (Nikon). For RI, RII and RIII spatial distance measurements, we calculated the centre-to-centre distances between the parent and child, specifically focusing on the child’s nearest parent. To measure the distances between the nuclear periphery and the centres of Foxo1 or Ebf1 probes67, we used DAPI as the reference point. The nuclei of individual cells were delineated through DAPI staining, and the measurements were based on the child (Foxo1 or Ebf1 probes) intersecting with the parent (DAPI-stained nuclei). For the distances between Foxo1 and γ-satellite signals, we measured the centre-to-centre distances between the parent and child, focusing on the child’s nearest parent. We selected the minimum distances for each parent since γ-satellite has multiple signals. While each parent can have more children, each child has only one parent for Nikon measurements.

Retroviral expression of Ebf1

Retrovirus was produced by transfection of Plat-E cells with retroviral plasmids containing Ebf1 cDNA and a selective green fluorescent protein (GFP) marker (pMYs-Ebf1-IRES-EGFP)42, or with GFP-only plasmids as the control (pMSCV-IRES-GFP, Addgene cat. no. 20672). Supernatants were collected at both 48 and 72 h post-transfection and subsequently concentrated using the Retro-X Concentrator (Takara, cat. no. 631456), following the manufacturer’s instructions. Young and old Rag2−/− CD19+ cells were cultured with 2 ng ml−1 of IL-7 for three days and then exposed to the resuspended virus with the supplement of 10 μg ml−1 polybrene. The cells were cultured for an additional 72 h and subsequently sorted on the basis of GFP expression before subsequent analysis.

Lentiviral expression of Wapl

Lentivirus was produced by transfection of 293T cells with the pHIV-Wapl-IRES-BFP and pHIV-IRES-BFP plasmids. To make these plasmids, the BFP DNA fragment was amplified from plasmid pEJS614_pTetR-P2A-BFPnls/sgNS (Addgene, cat. no. 108650), and then integrated into pHIV-IRES-Zsgreen (Addgene, cat. no. 18121) using the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, cat. no. E2621S) as per the manufacturer’s instructions, resulting in the generation of pHIV-IRES-BFP plasmids. Subsequently, Wapl cDNA was introduced into pHIV-IRES-BFP, creating the pHIV-Wapl-IRES-BFP. Supernatants were collected at both 48 and 72 h post-transfection and subsequently concentrated using the Lenti-X Concentrator (Takara, cat. no. 631232). D345 cells were infected with the resuspended lentivirus with the addition of 8 μg ml−1 polybrene. After infection, the cells were cultured for 72 h and sorted on the basis of BFP expression. The sorted cells were then cultured for 2–3 additional days to obtain an increased number of BFP-positive cells for experimental purpose.

H3K27ac HiChIP

The H3K27ac HiChIP procedure was executed employing the Arima Hi-C+ Kit (Arima Genetics Inc. cat. no. A101020), strictly adhering to the manufacturer’s guidelines outlined in the Arima-HiC+ documents A160168 v00 (HiChIP) and A160169 v00 (library preparation). In each sample, a total of 5 × 106 cells were utilized. These libraries were individually barcoded and then combined for sequencing on an Illumina NovaSeq instrument.

Intracellular staining of H3K27ac

Total bone marrow cells were obtained from young and old Rag2−/− mice to evaluate H3K27ac levels through intracellular staining. Cell surface markers were stained with PE anti-CD19 and BV421 anti-B220, washed twice with PBS and then stained in PBS with the eBioscience Fixable Viability Dye eFluor 780 (Invitrogen, cat. no. 65-0865-14). Following this, cells were washed with PBS and fixed and permeabilized using the Fix/Perm buffer provided by the Transcription Factor Buffer Set (BD, cat. no. 562725) according to the manufacturer’s instructions. Fixed and permeabilized cells were then washed twice with the Perm/Wash buffer form the Transcription Factor Buffer Set and stained with either anti-H3K27ac or IgG isotype control antibodies. Subsequently, the cells were washed three times with Perm/Wash buffer before being subjected to flow cytometry analysis for H3K27ac quantification.

Capture Hi-C

Whole-genome Hi-C libraries were first generated as previously described. To enrich the Igh locus (mm10, chr12: 113,201,001–116,030,000), SureSelect Target Enrichment probes (Supplementary Table 5) with 2× tiling density were designed and manufactured by Agilent (Agilent Technologies, Inc.). Hi-C libraries were hybridized to probes as specified by the manufacturer. Enriched libraries were sequenced using Illumina NextSeq sequencer to generate 2× 150 bp reads.

VDJ-seq

VDJ-seq was performed following the HTGTS-Rep-Seq protocol49,50. Briefly, 200 ng to 2 μg of genomic DNA isolated from primary pro-B cells sorted from young and old C57BL/6J mice was sonicated with Covaris using the setting of Durations 10 s, Peak power 50, Duty Factor 20%, Cycles/Burst 200. Sheared DNA was linearly amplified using a biotinylated JH1 primer and enriched with streptavidin C1 beads (Thermo Fisher Scientific). A bridge adaptor was ligated to enriched single stranded DNA to the 3′ end and then amplified by a nested primer set. Products were further amplified using P5–I5 and P7–I7 primers to be final libraries and sequenced by Illumina MiSeq sequencer to generate paired-end 250 bp reads. Primers or oligos used are listed in Supplementary Table 6.

Sequencing data bioinformatics analysis

For all analyses, we utilized the mouse reference genome mm10 and annotations from GENCODE (M21)68. The RNA-seq data and ChIP–seq data generated in this study were pre-processed from raw reads to expression level, differentially expressed genes analysis, mapped reads or tracks (bigWig files) following the procedures outlined in our previous studies69,70.

This manuscript incorporates heatmaps generated from pooled and down-sampled reads, ensuring equal representation from Hi-C or HiChIP data. Related statistical tests were performed with function in Python SciPy71 package. If not specially mentioned, most plots were generated by the matplotlib72 and seaborn73.

RNA-seq analysis

RNA-seq raw reads of young and old Rag2−/− primary pro-B cells were trimmed and mapped to the mouse reference genome (mm10) using STAR (v2.7.3a)74 and calculated the raw count using featureCounts package (gene-level) and fragments per kilobase of transcript per million mapped reads (or transcript per million) using RSEM package75.

For the rest of those RNA-seq experiments, raw reads were trimmed and mapped to mm10 by STAR (v2.7.3a)74 and quantified into reads per kilobase per million mapped reads with Cufflinks (v2.2.1)76. Bigwig tracks were also generated by STAR as signal quantified as reads per kilobase per million mapped reads. RNA-seq tracks were shown by IGV77.

ChIP–seq analysis

ChIP–seq raw reads were mapped to the mouse reference genome mm10 by Bowtie2 (v2.3.5)78. Only non-redundant reads with mapping quality score ≥10 were saved as BED files for the following analysis. Bigwig tracks were generated by bamCoverage in deepTools (v3.3.0)79 with parameters of --ignoreDuplicates --minMappingQuality 10 --normalizeUsing CPM for visualization and quantification aggregation analysis. Visualization of ChIP–seq tracks were performed by IGV77 or the cLoops2 plot module (v0.0.2)80. ChIP–seq peaks were identified using the cLoops2 callPeaks module. For H3K27me3, CTCF, Rad21, Brg1 and p300 ChIPs, the key parameters -eps 150 and -minPts 20,50 were employed. For H3K27ac data, the parameters -eps 150,300, -minPts 20,50 and -sen were used. Peak identification was performed on each replicate individually, followed by the compilation of a union set encompassing both conditions (young and old) and overlapping replicates. Differential peaks were identified for young and old conditions using a Poisson test. The test utilized average counts from replicates for each peak and considered normalized total reads of 30 million. A P value cut-off of <1 × 10−5 and a fold change cut-off of ≥2 were used to identify significant differential peaks. ChIP–seq peak or domain regions aggregation analyses, including heatmaps and average profile plots, were generated using the computeMatrix and plotHeatmap commands from the deepTools (v3.3.0) package. Quantification of ChIP–seq signals in Hi-C compartments was performed by the cLoops2 quant module. H3K27ac peak annotations were performed by the anoPeaks.py script in cLoops2 package, which is available at https://github.com/YaqiangCao/cLoops2/blob/master/scripts/anoPeaks.py. H3K27ac peaks within a 2-kb range upstream or downstream of the transcription start site (TSS) are annotated as promoters, while those outside this range are labelled as enhancers for the downstream analysis.

Hi-C, H3K27ac HiChIP and capture Hi-C analysis

Hi-C, H3K27ac HiChIP and capture Hi-C raw reads were processed to the mouse reference genome mm10 by HiCUP (v0.7.2)81 with the settings of Arima. High-quality and unique paired-end tags (PETs) from HiCUP were further processed to the HIC file through Juicer (v1.6.0)82 for visualization with Juicebox83. These PETs were also processed with cLoops2 (v0.0.2)80 for quantifications. Chromosome X was excluded from the analysis. Replicates from H3K27ac HiChIP samples were combined and down-sampled equally to 39 million PETs for all following analysis.

Hi-C compartment analysis of eigenvectors first principal component (PC1) was obtained by hicPCA in HiCExplorer3 package (v3.6)84 with parameters of -noe 1 at the resolution of 100 kb. We employed a two-pass Mahalanobis distance (MD) calculation, an effective method for outlier detection based on data point distribution, along with a chi-squared test for Hi-C compartment PC1 at the 100-kb bin level to detect PC1 sign flip bins. This approach aligns with the strategy employed in the recent study dcHiC85. The MD is calculated as \(}=}((X-_)\times ^\times _\right)}^)\), where X is the matrix for PC1 obtained from the Hi-C compartment analysis. Each row represents a bin, and each column represents a sample (young or old). C−1 is the inverse covariance matrix of X, Xc is vector of row-wise mean of X and diag is the function used to extract a diagonal array from a matrix. We performed the chi-squared test with the MD, and a P value cut-off of 0.01 was set as the significance cut-off for detecting outliers. For the first pass calculation of MD, outliers were detected with all bins. For the second pass calculation, outliers were removed for calculation Xc and C−1. Then, MD distances were calculated for all bins on the basis of the first pass-outlier removed Xc and C−1. The code for the analysis to generate the MD distances and P values, the plot, significant changed bins and associated genes is implemented as the comparComp.py script in the cLoops2 package and is available at https://github.com/YaqiangCao/cLoops2/blob/master/scripts/compareComp.py.

Hi-C TADs were called by Juicer arrowhead with parameters of -r 25000 -k KR. We employed ESs to quantify TAD interaction strength. The ES is calculated by dividing the number of PETs interacting exclusively within a TAD by the number of PETs with one end within the TAD. Like the comparison of compartments, we implemented the two-passes MD calculation and chi-squared test to get the significant changed TADs. The MD is calculated as \(}=}((X-_)\times ^\times _\right)}^)\), where X is the matrix for TAD ESs. Each row represents a TAD, and each column represents a sample (young or old). C−1 is the inverse covariance matrix of X, Xc is vector of row-wise mean of X, and diag is the function used to etract a diagonal array from a matrix. We performed the chi-squared test with the MD, and a P value cut-off of 0.01 was set as the significance cut-off for detecting outliers. For the first pass caculation of MD, outliers were detected with all TADs. For the second pass caculation, outliers were removed for calculation Xc and C−1. Then, MDs were caculated for all bins based on the first pass-outlier removed Xc and C−1. The TAD ES was calculated by dividing the number of intra PETs (PETs with both ends located in the TAD) by the number of inter PETs (PETs with only one end located in the TAD). We used the ES as a metric because we found that the IgH domain had the highest change, along with domains contain Ebf1 and Pax5 in the top. The code for the analysis to generate the MD distances and P values for the TADs, the plot, significantly changed TADs and associated genes is implemented as the compareDom.py script in the cLoops2 package and is available at https://github.com/YaqiangCao/cLoops2/blob/master/scripts/compareDom.py.

H3K27ac HiChIP data viewpoints analysis was used to examine the changes in interactions for H3K27ac ChIP–seq peaks. For each viewpoint (H3K27ac peak), an ES was calculated using the H3K27ac HiChIP contact matrix at a 1-kb resolution. The ES was centred on the peak and incorporated 100-kb upstream and downstream regions as the background. A higher ES indicates stronger interactions originating from the viewpoint. To visually emphasize the signals, the upper right corner matrices of the ±100-kb contact matrices centred around each viewpoint (H3K27ac peak) were subjected to two rounds of log10 transformation and presented as heatmaps. cLoops2 plot module was utilized to generate visualizations of HiChIP data, including one-dimensional signal profiles, ChIP–seq tracks, arches depicting the number of PETs for combinations of H3K27ac peaks and heatmaps or scatter plots.

Virtual 4C analysis

Virtual 4C analysis was performed using Cooler2 to extract all contacts involving the 3′ CBE from the Hi-C contact maps86. The bin size used for this analysis was 5,000 bp and ‘VC_SQRT’ was used as the Hi-C normalization method. The bin used for the 3′ CBE bait was chr12:113220000-113225000 (mm10).

VDJ-seq analysis

VDJ-seq analysis was performed using HTGTS-Rep49,50. This computational pipeline determined the frequency of V and D gene usage for each HTGTS experiment. This pipeline was also used to determine the distribution of CDR3 lengths and the percent of productive versus non-productive recombination events.

Statistics and reproducibility

The presentation of data and the statistical tests utilized are detailed in each figure legend. For datasets with a sample size (n) less than three, individual data points are displayed, assuming a normal data distribution, but this was not formally tested. Unless otherwise specified, the statistical tests were performed using GraphPad Prism v.9.5.0. For Fig. 4f, all P values were reported as 0.000000 by scipy.stats.wilcoxon. No statistical method was used to predetermine sample size, but our sample sizes are similar to those reported in previous publications54. No data were excluded from the analysis. All biologically independent replicates are explicitly identified in the figure legends. The gating strategy of flow cytometry is shown in Supplementary Fig. 1. Fluorescence-activated cell sorting data and microscopy (FISH) images are representative of at least three independent biological replicates. Experiments were not randomized, and data collection and analysis were not performed blind to experimental conditions.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.