Temporal analyses reveal a pivotal role for sense and antisense enhancer RNAs in coordinate immunoglobulin lambda locus activation

Abstract Transcription enhancers are essential activators of V(D)J recombination that orchestrate non-coding transcription through complementary, unrearranged gene segments. How transcription is coordinately increased at spatially distinct promoters, however, remains poorly understood. Using the murine immunoglobulin lambda (Igλ) locus as model, we find that three enhancer-like elements in the 3′ Igλ domain, Eλ3–1, HSCλ1 and HSE-1, show strikingly similar transcription factor binding dynamics and close spatial proximity, suggesting that they form an active enhancer hub. Temporal analyses show coordinate recruitment of complementary V and J gene segments to this hub, with comparable transcription factor binding dynamics to that at enhancers. We find further that E2A, p300, Mediator and Integrator bind to enhancers as early events, whereas YY1 recruitment and eRNA synthesis occur later, corresponding to transcription activation. Remarkably, the interplay between sense and antisense enhancer RNA is central to both active enhancer hub formation and coordinate Igλ transcription: Antisense Eλ3–1 eRNA represses Igλ activation whereas temporal analyses demonstrate that accumulating levels of sense eRNA boost YY1 recruitment to stabilise enhancer hub/promoter interactions and lead to coordinate transcription activation. These studies therefore demonstrate for the first time a critical role for threshold levels of sense versus antisense eRNA in locus activation.


INTRODUCTION
The spatiotemporal control of gene transcription is a highly intricate and tightly regulated process that is crucial for eukaryotic de v elopment.Gene transcription r equir es r egulatory e v ents at promoters, where transcription factors bind to specific motifs that lie upstream of the transcription start site (TSS), to acti vate assemb ly of the RNA polymerase II (RNAPII) pre-initiation complex ( 1 ).Whilst these promoter-specific e v ents are important to control basal transcription activity, much greater regulation stems from a second, more abundant class of regulatory element, namely transcription enhancers ( 2 ).These can reside many thousands of bases from the cognate gene promoters, either upstr eam or downstr eam, and ar e composed of concentrated clusters of recognition motifs for di v erse transcription factors, often including nucleosome-binding factors, architecture factors and transcription coactivators ( 3 ).Transcription enhancers physically interact with their target gene promoters to vastly increase the le v el at which the gene is transcribed ( 4 ).Functional enhancer-promoter contacts are strongly influenced by the way in which chromosomes are folded in three-dimensional space ( 5 ); the latter occurs in a hierarchical manner to gi v e compartments, topologically associating domains (TADs) and insulated neighbourhood domains (INDs), that are thought to represent structural and functional units of genome organization.Physical contacts between different cis-acting elements across structural unit boundaries are relati v ely infrequent whereas efficient tissue-specific gene expression requires transcriptional enhancers and their cognate promoters to be constrained within the same genome structural unit ( 6 ).
Antigen receptor loci are essential to generate a highly div erse adapti v e immune system.These loci, howe v er, present a unique problem for enhancer-mediated gene activation: Generation of antigen receptor di v ersity requires recombination between complementary gene segments that can be many kilobases to megabases apart.These gene segments must be coordinately activated via non-coding transcription to increase their accessibility prior to recombination ( 7 , 8 ).Enhancers are central to regulating this non-coding transcription ( 9 ) but how enhancers coordinately activate promoters that are far apart in the primary sequence, is poorly understood.This problem is exacerbated by the presence of up to 100 gene segments and many potential regulatory elements in some loci.Since the appropriate chromatin environment is a pr er equisite to facilitate enhancerpromoter interactions, initial studies focused on chromatin folding of the IgH and Ig loci, using DNA fluorescence in situ hybridization (FISH) and 3C deri vati v e technologies (10)(11)(12)(13)(14). From this, it was proposed that prior to V(D)J r ecombination, antigen r eceptor loci form a poised state where they are contracted via a series of loop domains.Contraction is tightly associated with binding of histone modifiers, lymphocyte-specific transcription factors and architecture factors, such as p300, IRF4, PAX5, E2A, CTCF, cohesin and YY1 at interspersed DNA regulatory elements throughout the locus and correlates with enhanced noncoding transcription of unrearranged gene segments (15)(16)(17)(18)(19)(20).Howe v er, these studies did not explore the regulation of antigen receptor locus activation and chromatin folding in fine detail.Indeed, w hilst anal ysis of chromatin folding in B-cells at different stages of de v elopment enab les predictions regarding the coordination of events, these studies cannot truly unravel the temporal order of locus specific enhancer-promoter communications and coordinated activation in any detail.
A barrier to the temporal analysis of coordinate locus activation has been the lack of a homogenous population of lymphocytes in which antigen receptor locus activation can be induced.Activation of light chain loci is a hallmark of the pro-B to pre-B transition and previous studies showed that Ig locus activation absolutely relies on the E 3-1 enhancer ( 21 ).Notably, this enhancer contains binding sites for the transcription factor, IRF4 and we showed that remar kab ly, equipping with pro-B cells with pre-B le v els of just this single transcription factor, is sufficient to completely activate transcription and recombination of unrearranged Ig gene segments ( 7 ).This, together with the small size of the murine Ig locus, spanning just ∼230 kb, and low number of functional gene segments provides an excellent system to tem-porally dissect locus activation.Ther efor e, to unravel the regulation of enhancer-promoter interactions and changes in chroma tin organiza tion r equir ed for coordinate V and J gene segment activation, we developed transgenic mice and a pro-B cell line that expresses an inducible IRF4.By studying the dynamics of transcription factor recruitment and changes in Ig chroma tin organiza tion, we built a detailed picture of the stages of activation and show that coordinate transcription factor binding to three enhancer-like elements is essential to form an acti v e enhancer hub.This hub then coor dinately acti vates transcription through V and J gene segments.Remar kab ly, the interplay between sense and antisense enhancer RNAs (eRNAs) is central Ig activation: Threshold le v els of sense v ersus antisense eRNA are vital to control YY1 recruitment, stabilisation of enhancer hub formation and enhancer-promoter interactions, and lead to high le v els of Ig non-coding transcription.
Non-transgenic mice (CBA / C57BL / 6J) were obtained from the Uni v ersity of Leeds animal facility.IRF4-ER transgenic mice were generated in the same way as the PIP2, PIP3 and PIP4 transgenic mice described previously ( 7 ) where Irf4 was expressed under control of the pro-B cell specific 5 promoter and LCR.Here, IRF4 was fused to the estrogen receptor hormone binding domain and the fusion gene substituted for Irf4 in the 5 promoter / LCR cassette.Animals were sacrificed at 5-7 weeks, bone marrow was removed from femurs and used for the isolation of pro-or pre-B cells by flow cytometry.Equivalent numbers of male and female animals were used overall.All animal procedures were performed under Home Office licence PPL 70 / 7697 and P3ED6C7F8, following re vie ws by the Uni v ersity of Leeds ethics committee.They were housed in a full barrier facility, with no more than six animals per cage, where all mice are free of common pathogens, including murine norovirus, P asteur ella and Helicobacter.
HEK293T were a kind gift fr om Pr of.Mark Harris and Phoenix cells were generously supplied by Dr Garry Nolan.They were maintained in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% foetal calf serum, 4 mM L -glutamine, 50 U / ml penicillin and 50 g / ml str eptomycin.Cells wer e grown in a humified incuba tor a t 37 • C with 5% CO 2 .
Pro-B cells were flushed from the femurs of 5-7 week old mice and cultured in pro-B cell medium as described previously ( 7 ).Primary cells wer e cultur ed at 33 • C, 5% CO 2 for 7 days with the addition of 5 ml fresh medium after 4 days.

Generation of A-MuLV-transformed pro-B cell lines
The AB010 cell line ( 23 ), which secretes Abelson murine leukaemia virus (A-MuLV), was grown for two days past confluency in supplemented DMEM.The virus containing supernatant was removed and concentrated using a Centricon Plus-70 centrifugal device.Bone marrow was flushed from the femurs of 5-7-week-old mice and cells were immediately infected with A-MuLV.Red blood cells were lysed for ten min by suspension in 168 mM NH 4 Cl.Infection with A-MuLV was performed by the addition of 1 ml of primary cells at a concentration of 2 × 10 6 cells / ml to 1 ml of concentrated viral supernatant, in the presence of 8 g Polybr ene (Millipor e).Cells wer e incuba ted a t 37 • C for 2.5 h with agitation e v ery 20 min and plated at concentrations of 1 × 10 6 cells / ml in semi-solid agar (RPMI supplemented with 20% foetal calf serum, 2 mM L -glutamine, 50 g / ml streptomycin, 50 U / ml penicillin, 50 M ␤mercaptoethanol and 0.3% bacterial agar (Oxoid Ltd).Following infection, cells were maintained in a humidified atmosphere a t 37 • C , adding 1 ml of semi-solid agar e v ery 4 days.

Generation of MSCV-IRF4-ERT2 cell lines
Retr oviruses, pr oduced using the MSCV-IRF4-ERT2-IRES-GFP construct, were transduced into the A-MuLV immortalized pro-B cell line.Infection was monitored via GFP expression and flow cytometry.To generate monoclonal cell lines, 1 × 10 4 cells expressing the highest le v el of GFP were purified by flow cytometry and plated in 10 ml of semi-solid agar.After 10 days, macroscopic colonies wer e transferr ed to RPMI in 24-well plates and expanded.

Tamoxifen and Imatinib treatment of cell lines
The IRF4-ERT2 protein was activated in PIPER-15 cells by addition of tamoxifen.Inductions were performed by resuspending 1-5 × 10 6 cells at 0.5 × 10 5 cells / ml in RPMI; 4hydro xytamo xifen (Insight Biotechnology; HY-16950) was added to a final concentration of 2 M. Cells were incuba ted a t 37 • C with 5% CO 2 for the times indica ted.Imatinib was added to PIPER-15 cells, resuspended as above, a t final concentra tions of 1-100 nM.Cells were incubated for 48 h prior to harvest.
Nuclear extracts were prepared by resuspending PBS washed cells at a density of 1 × 10 6 cells / ml in 1 ml of lysis buffer (10 mM Tris pH 8, 10 mM NaCl, 0.2% NP-40, 50 g / ml PMSF, 1 × Complete ™ protease inhibitor cocktail, Roche) and incubating on ice for 20 min.Nuclei were pelleted at 800 g for 2 min before resuspending in 100 l 1 × Laemmli loading buffer and boiling for 5 min.
Whole cell and nuclear extracts were either used immediately for western blotting or flash frozen on dry ice and stored at -80 • C until required.

Western blotting
Following electr ophoresis, pr oteins wer e transferr ed to PVDF membrane (Immobilon-P, IPVH00010, Millipore) in a Trans-Blot Turbo transfer system (Bio-Rad) for 30 min at 25 V.The PVDF membrane was blocked in a solution of 5% non-fat milk po w der in TBS-T (50 mM Tris pH 7.6, 150 mM NaCl, 5% milk, 0.05% Tween-20) for 1 h at room temperature.All primary antibody hybridisations were conducted overnight at 4 • C, whereas secondary or tertiary antibod y hybridisa tions were performed a t room tempera ture for an hour.Antibodies are gi v en in Supplementary Table 1 and were used at the dilutions recommended by the manufacturer.After each hybridisation, membranes were washed with TBS-T, with changes e v ery 5 min for 1 h.Membranes were de v eloped by incubation with enhanced chemiluminescence substrate (Thermo Scientific) for 2 min at room temperature and imaged using a G:BOX ChemiXT4 system (Syngene).

Total RNA extraction and reverse transcription
Total RNA was extracted from approximately 2 × 10 6 cells using TRIzol (Invitrogen #3289) according to the manufacturer's instructions, followed by treatment with 2 U DNase I (Worthington) for 1 hr at 37 • C in 100 l of 1 x NEB DNase I buffer (10 mM Tris pH 7.5, 2.5 mM MgCl 2 , 0.5 mM CaCl 2 ).Following phenol-chloroform extraction and ethanol precipita tion, RNA concentra tion was determined using a DS11 + spectrometer (DeNovix).
1 g of RNA was re v erse transcribed with M-MuLV re v erse transcriptase (Invitro gen).Briefly, 1 g of RN A was added to 2.5 M oligo dT primer (or strand-specific primer, where noted), 500 M dNTPs and ddH 2 O to gi v e a total volume of 12 l.This was incuba ted a t 65 • C for 5 min and immediately placed on ice before addition of 4 l first strand buffer (Invitrogen), 10 mM DTT and 1 l RNasinPlus (Promega).The reaction was incuba ted a t 37 • C for 2 min, followed by addition of 1 l Moloney-Murine Leukaemia Virus Re v erse Transcriptase (Invitrogen), incuba tion a t 37 • C for 50 min prior to hea t inactiva tion a t 70 • C for 15 min.

Real-time PCR using SYBR Green
Quantitati v e PCR was performed using a Corbett Rotor-Gene 6000 machine and analysed using the Corbett Rotor-Gene 6000 Series Software (v.1.7,build 87).A typical qPCR reaction contained 5 l 2 × SensiFAST SYBR No-Rox mix (Bioline #BIO-98080), 2-10 ng DN A template, or cDN A at a final dilution of 1:100, 400 nM of each primer in a total volume of 10 l.Primer sequences are gi v en in Supplementary Table 1.All reactions were performed in duplicate.In each case, a standard curve of the amplicon was analysed concurrently to evaluate the amplification efficiency and to calculate the relati v e amount of amplicon in unknown samples.R 2 values were 1 ± 0.1.A typical cycle consisted of: 95 • C for 3 min, followed by 40 cycles of 95 • C for 5 s, Tm for 10 s and 72 • C for 10 s, where T m = melting temperature of the primers.A melt curve, to determine amplicon purity, was produced by analysis of fluorescence as the temperatur e was incr eased from 72 • C to 95 • C. Amplicons were 100-200 bp.

Analysis of V 1-J 1 recombination
Primary pro-B cell cultures from IRF4-ER transgenic mice were expanded for seven days, as described ( 7 ).Tamoxifen was added to at a final concentration of 2 M for the induction times indicated, prior to cell harvest.Pro-B cells were then purified by flow cytometry with 2 M Tamoxifen present in all buffers and DNA was pr epar ed as described ( 7 ), using at least four phenol / chloroform extractions to remove contaminants prior to ethanol precipitation.The resuspended DNA was quantified using a Quant-iT ™ PicoGreen ™ assa y (In vitrogen), according to the manufacturer's instructions.DNA amounts were further normalised using 2-3 ng in qPCR reactions and Intgene III primers.V 1 / J 1 recombination was determined via nested qPCR, using 3 ng DNA and 15 cycles in the first round of PCR.Following a 10-fold dilution of the product, 1.5 l was used in the second round qPCR reaction.Primer sequences are gi v en in Supplementary Table 1.

T r ansfection of HEK293T and Phoenix cells
Transfection of HEK293T and Phoenix cells was carried out using PEI (Alfa Aesar #043896.01).Twenty-four hours before transfection, 3 × 10 6 cells were plated per 10 cm dish in complete DMEM.Three hours prior to transfection, the medium was changed to fresh complete DMEM medium.Plasmid DNA (10 g) was mixed well with 500 l Opti-MEM ™ by gentle vortexing.Concomitantly, 30 l of PEI solution (1 mg / ml) was diluted with 500 l of OptiMEM medium.The solutions were then mixed well for 15 s, followed by incubation at room temperature for 15 min.The mixture was added to cells dropwise; cells were then incuba ted a t 37 • C for 48 h prior to harvest.

T r ansfection of 103 / BCL-2 cells
Electroporation was carried out using the Nucleofector ™ Kit (LONZA # VPA-1010) according to manufacturer's instructions.Briefly, 4 × 10 6 cells were washed twice with ice cold PBS and resuspended in 100 l of transfection reagent (82 l nucleofector plus 12 l supplement 2), followed by addition of 2 g plasmid DNA.Cells were then transferred to a cuvette and electroporated at setting Z01 of the AMAXA electroporator.Following addition of 500 l complete RPMI medium, cells were decanted into a 6-well plate using a sterile pastette; an additional 1400 l of RPMI medium was added, followed by incubation at 33 • C overnight.Twenty hours prior to harvest, cells were temperature shifted to 39.5 • C to inactivate the temperaturesensiti v e v-Ab l kinase ( 24 ) and trigger light chain transcription.

Lucifer ase r eporter assay
The luciferase assay was carried out using the Dual-Luciferase Kit (Promega) according to manufacturer's instructions.Cells were washed twice with ice cold PBS and resuspended in 1 ml Passi v e Lysis Buffer, followed by gentle shaking at room temperature for 15 min.Following transfer to a fresh Eppendorf tube, the lysate was vortexed vigorously for 15 s and centrifuged at 16 000 g for 10 min at 4 • C. 100 l of the Luciferase Assay substrate was pre-dispensed into a luminometer tube.20 l of the lysate was added, followed by determination of firefly luciferase activity using the SIRIUS luminometer v3.0.Renilla luciferase activity was measured by addition of 100 l of Stop & Glo ™ reagent.

Flow cytometry
Primary pro-B and pr e-B cells wer e stained with FITC and PE conjugated antibodies as described ( 7 ).Antibody labelled cells or GFP expressing cells were purified by flow cytometry using a FACSMelody ™ cell sorter (Becton Dickinson).GFP expr essing cells wer e analysed by flow cytometry using a CytoFLEX flow cytometer (Beckman Coulter, USA) to determine the percentage of cells that had successfully been transduced.Cells wer e pr epar ed for flow cytometry by washing with, and resuspension in, ice cold PBS.

Production of r etro vir al particles
Retroviral particles were generated using Phoenix cells ( 25 ).Twenty-f our hours bef ore transfection, 3 × 10 6 Phoenix cells were plated per 10 cm dish in complete DMEM.Three hours prior to transfection, the medium was changed to DMEM supplemented with 5% foetal calf serum, 4 mM Lglutamine. 4 g of MSCV-IRF4-ERT2-GFP construct was mixed with 500 l of OptiMEM by gentle vortexing.Concomitantly, 12 l of PEI (1 mg / ml) was diluted with 500 l of OptiMEM.The solutions were then mixed with gentle vortexing for 15 s, followed by incuba tion a t room temperature for 15 min and dropwise addition to cells.Cells were incuba ted a t 37 • C for 48 and 72 h prior to harvest.The retrovirus-containing supernatant was filtered through a 0.45 m syringe filter, flash frozen on dry ice and stored at -80 • C until use.

Production of lentiviral particles
Lentiviral particles were produced in HEK293T cells by transfection with the lentiviral backbone constructs, packaging construct (pCMVR8.74) and envelope construct (pMD2.G).For lentiviral backbone constructs, pLKO.1-TRC was used to produce shRNA-mediated knock-down lentiviral particles.3 × 10 6 HEK293T cells were plated per 10 cm dish in complete DMEM 24 h before transfection.Three hours prior to transfection, the medium was changed to DMEM supplemented with 5% foetal calf serum, 4 mM L -glutamine .Separately, 4.9 g of pLKO .1 shRNA plasmid, 2.6 g of pCMVR8.74 and 2.5 g of pMD2.G were mixed with 500 l of OptiMEM medium by gentle vortexing, whereas 30 l of PEI stock solution (1 mg / ml) was diluted with 500 l of OptiMEM medium.Transfection, harvest and storage of lentiviruses was then as described for retro viruses abo ve.

Knockdown of Med23 , Med1 , Yy1, Spi1 and eRNAs
pLKO.1, expressing the a ppropriate shRN A, was cotransfected into HEK293T cells with the packaging plasmids, pCMVR8.74and pMD2.G, to produce lentiviral particles.The resulting lentivirus was used to transduce PIPER-15 cells by spin-fection via centrifugation at 800 g for 30 min a t 32 • C .After 48 h, puromycin (Cayman Chemical) was added at a final concentration of 2 g / ml, followed by incubation at 37 • C for 7 days.

Knockout of E 3-1 and HSE-1 enhancers
Two CRISPR sgRNA-specifying oligonucleotides that flank the PU.1 / IRF4 sites in each enhancer element (E 3-1 and HSE-1) were designed as above.E 3-1 gRNA1 / HSE-1 gRNA1 oligonucleotides were annealed and cloned into lenti-CRISPR v2 whereas E 3-1 gRNA2 and HSE-1 gRNA2 oligonucleotides were cloned into lenti-sgRNA-MS2-zeo.Lentiviral production was performed as described above.Transductions of PIPER-15 cells were performed in a sequential manner.5 × 10 5 PIPER-15 cells were spin-fected with 500 l of E 3-1 gRNA2 or HSE-1 gRNA2 lentiviruses; transduced cells were selected with 100 g / ml Zeocin (Alfa Aesar J67140) after 48 h.After one week of selection, cells were spin-fected with E 3-1 gRNA1 or HSE-1 gRNA1 lentivirus and selected for one week with 0.25 g / ml puromycin.Monoclonal cell lines were generated using semi-solid agar and clones were screened for knockouts by PCR using the primers HSE-1delF / R and E 3-1del F / R (Supplementary Table 1).Monoclonal cell lines with apparent deletions in these regions were amplified using the above primers; the products were cloned and knockout of the respecti v e region confirmed by Sanger sequencing.

Chromatin Immunoprecipitation (ChIP)
ChIP in primary pro-and pre-B cells was carried out according to Boyd and Farnham ( 26 ) with modifications using 2 × 10 7 cells per e xperiment.ChIP e xperiments in PIPER-15 cells and with anti-E2A and anti-MED1 antibodies in primary pro-and pr e-B cells, wer e performed according to Nowak et al. ( 27 ) by first cross-linking with 2 mM Disuccinimid yl Glutara te (DSG, Sigma 80424) and then with 1% formaldehyde.The antibodies and dilutions used are gi v en in Supplementary Table 1.The r ecover ed DNA was analysed using quantitati v e PCR and the primers shown in Supplementary Table 1.

Chromatin conformation capture (3C)
3C was carried out according to Dekker et al ( 28 ) with modifications. 1 × 10 7 PIPER-15 cells were used per experiment and following preparation of cross-linked nuclei, samples were flash frozen in liquid nitrogen and stored a t -80 • C .Stored nuclei were resuspended in 500 l 1.2 × NEB Dpn II buffer (50 mM Bis-Tris-HCl pH 6.0, 100 mM NaCl, 10 mM MgCl 2 , 1 mM DTT) in a screw capped tube.SDS was added to a final concentration of 0.3% followed by vigorous pipetting.The nuclei were shaken at 200 rpm for 60 min at 37 • C with pipetting e v ery 15 min, to pre v ent aggregation, prior to addition of Triton X-100 to a final concentration of 3%, and incubation at 37 • C for 60 min with shaking.The nuclei were then digested by addition of 100 units of Dpn II (NEB, R0543M) and incubation at 37 • C for 4 h with shaking, followed by an overnight digestion with an additional 100 units of Dpn II.Following addition of a further 100 units of Dpn II and incubation for 4 h at 37 • C, the restriction enzyme was inactivated by incuba tion a t 65 • C for 20 min, and digested nuclei transferred to a fresh tube.Ligation was performed in 7 ml of 1 × ligase buffer (50 mM Tris-HCl pH 7.5, 10 mM MgCl 2 , 1 mM ATP, 5 mM DTT) with 25 U T4 DNA ligase (Roche) at 16 • C overnight.RNase A was then added to a final concentration of 10 g / ml at 37 • C for 30 min; crosslinks were re v ersed by addition of proteinase K to a final concentration of 100 g / ml and incubation at 65 • C for at least 4 h.Ligated DNA sample was phenol / chloroform extracted, precipitated with ethanol, and resuspended in 100 l of TE.

Pr epar ation of BAC template for 3C analysis
Bacterial artificial chromosome (BAC) Rp23-24i11 was obtained from Children's Hospital Oakland Research Institute and contains the 3 half of the murine Ig locus.Dpn II (NEB, R0543M) is blocked by Dam methylation; therefore, BAC DNA was digested with its isoschizomer Sau3AI (NEB, R0169S) and ligated at high concentration to generate all possible ligation products as a 3C normalisation control.20 g of BAC DNA was treated with 25 U of Sau3AI in a total volume of 500 l at 37 • C overnight.The digested BAC DNA was phenol-chloroform e xtracted, recov ered by ethanol precipitation and resuspended in 40 l TE.BAC DNA was ligated with 2000 cohesi v e end units / ml of T4 DNA ligase in a total volume of 60 l at 16 • C overnight.The ligated products were phenol / chloroform extracted, ethanol precipitated and resuspended in 100 l of TE.
Nested PCR assay to detect 3C interactions E 3-1 was used as a viewpoint to determine interactions within the Ig locus.A nested PCR assay was used to detect 3C interactions between E 3-1, HSE-1 and other cisacting elements using the primers gi v en in Supplementary Table 1.Nested PCR reactions were also performed on the BAC control template to correct for differences in primer efficiency.The first round of PCR was performed using Taq DN A pol ymerase.For the second round, TaqMan qPCR was conducted in duplicate in 10 l final volume with 5 l of 1:10 diluted first round PCR product, 400 pM each primer, 100 pM 5 nuclease probe and 5 l qPCRBIO probe mix (PCRBIO PB20.21-05).For Supplementary Figures 1A, B and 4G, H, only a single round of qPCR was performed, using TaqMan probes and the primers gi v en in Supplementary Table 1; HSE.1 was used as an additional viewpoint in Supplementary Figures 1B and 4G.All 3C samples were normalised by analysis of interactions in the Ercc3 locus which is expected to be consistent across all cell types ( 29 ).

In vitro transcription of enhancer RNAs
pLK O-T7p-sE 3-1e, pLK O-T7p-asE 3-1e and pLKO-T7p-randomRNA were linearized with EcoRI which cleaves just downstream of the respecti v e eRNA sequences.These were used as templates for in vitro transcription of sE 3-1e, asE 3-1e and random RNAs with T7 RNA polymerase (NEB, M0251S), according to the manufacturer's instructions.The in vitro transcribed products wer e tr eated with DNaseI to digest the template DNAs, ethanol precipitated and resuspended in DEPC-treated deionized water.

Electrophoresis of enhancer RNAs
Agarose gel electrophoresis of enhancer RNAs was conducted as described previously ( 30 ).Briefly, 1 g of enhancer RNA was heated at 95 • C for 2 min and placed on ice for 2 min.RNAs were incubated at 37 • C for 2 h and then mixed with nati v e loading buffer (10 × stock: 15% ficoll, 0.25% bromophenol blue, and 0.25% xylene cyanol FF) before loading onto a 1% agarose gel in TAE.Electrophoresis was at 40 V for 1.5 h at 4 • C.

RNA immunoprecipitation (RIP)
RIP was performed according to ( 31 ).The IgG and YY1 antibodies used are gi v en in Supplementary Table 1.The r ecover ed RNA was reversed transcribed with strand specific primers and then analysed using quantitati v e PCR and the primers shown in Supplementary Table 1.

A T A C-seq
ATAC-seq was performed as described previously ( 32 ) with minor modifications.Briefly, 5 × 10 4 cells were pelleted at 300 g for 5 min, washed with 50 l PBS and pelleted for 5 min at 300 g.Cells were lysed by resuspension in 50 l of ATAC-seq RSB (10 mM Tris-HCl pH 7.4, 10 mM NaCl and 3 mM MgCl 2 ) containing 0.1% NP40, 0.1% Tween-20 and 0.01% digitonin and incubation on ice for 3 min.Nuclei were washed to remove contaminating mitochondria with 1 ml of RSB containing 0.1% Tween-20 and pelleted at 500 g for 10 min.Nuclei were then resuspended in 50 l of transposition mix (25 l 2 × TD buffer, 2.5 l transposase, 16.5 l PBS, 0.5 l 1% digitonin, 0.5 l 10% Tween-20 and 5 l water) and incubated on a thermomixer at 37 • C for 30 min at 900 rpm.Reactions were purified using a Qiagen MinElute PCR-purification column.Library preparation was performed as described previously ( 33 ) with 10 cycles of amplification and purification using a Qiagen MinElute PCR-purification column.Samples were paired-end sequenced by Novogene on a NovaSeq 6000 S4 flow cell with a read length of 150 bp.

Analysis of next generation sequencing data
Accession numbers of all datasets used, are gi v en in Supplementary Table 1.

ChIPseq
Read files in FASTQ format were downloaded from the European Nucleotide Archi v e (ENA; https://www.ebi.ac.uk/ena ) and sequencing adapters were removed by Trim-Galor e (0.5.0).Reads wer e aligned to the Mus musculus (mm9) genome using Bowtie2 (2.3.4.2) and default parameters; m ultima pping reads as well as poor quality alignments wer e r emoved using Samtools (1.9).Peaks wer e called using MACS2 (2.1.0),for transcription factors, using default parameters.Visualisation was performed using the Integrated Genome Browser IGV (2.4.2) after converting the bedgraph output from MACS2 into a binary 'tiled' format using IGV tools (2.3.98).

A T A C-seq
Read files were downloaded from the ENA, trimmed and aligned as above, the Bowtie2 (2.3.4.2) max insert parameter (-X) was set to 2000 to enable the mapping of large inserts that are typical of ATAC-seq.Multimapping reads wer e r emoved by Samtools (1.9) befor e peak calling.ATACseq peaks were called by MACS2 (2.1.0)with the parameters -nomodel -shift 150 -extsize 300.
For new ATAC-seq data: Following quality checking of FASTQ files by FastQC v0.12.1, r eads wer e trimmed and aligned as above.PCR duplicates wer e r emoved using Picar d Mar kDuplicates v3.0.0.Prob lematic genomic regions present in the ENCODE Blacklist ( 34 ) were removed from the aligned files and further quality control of the aligned files was performed using Samtools v1.17.The deep learning based peak caller LanceOtron v1.0.8 (with a peak score cut-off value of 0.5) was used to call peaks.BigWigs were generated using the deepTools (v3.5.1) bamCoverage command, with the flags -extendReads -normalizeUsing RPKM, and visualized in the UCSC genome browser.

Hi-C
Read files (FASTQ) were downloaded and trimmed as above, before being aligned separately to the mm9 genome using Bowtie2 (2.3.4.2).The HOMER program (4.9) make-TagDirectory was used to process the aligned reads into a tag directory for downstream analysis.Significant interactions occurring in the Ig locus were identified with the HOMER script analyzeHiC.This command was run with the following parameters: -res 10000 -interactions < interaction file > -pos < region of interest > -center.This script identifies and reports pairs of regions that have a significantly increased number of interactions than would be expected from the background model.The 'center' argument r e-centr es the r egions outputted to the average of the position of the Hi-C reads participating in the interaction.Visualisation of the Hi-C interactions was performed using Circos (0.69).

Statistical analyses
Statistical analyses were performed using GraphPad Prism v9.Analyses of fold changes between biological replicates, using biolo gicall y distinct samples from the same types of cells, were performed using a paired Student's t test where * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

PU.1 and IRF4 binding to the E 3-1 enhancer activates Ig gene transcription
The relati v el y simple organisation of the small m urine lambda light chain locus offers an excellent system to dissect temporal and coordinate activation of antigen receptor loci.This locus is thought to have arisen by an evolutionary duplication e v ent ( 35 ), resulting in similarly organised 5 and 3 domains, each with only 3-4 gene segments and several DNA regulatory elements (Figure 1 A).J ust lik e other antigen receptor loci, the V and J gene segments are many kb apart but crucially, ∼70% of Ig recombination occurs between V 1 and J 1 ( 36 ).Ther efor e, the mechanism of coordinate gene segment activation can be investigated by focusing on just these two gene segments.Recombination requires non-coding transcription through the unrearranged gene segments and the B cell specific enhancer, E 3-1, is pivotal to this regulation ( 21 , 37 ).Consistent with this, the significant increase in V 1 and J 1 transcription from proto pre-B cells (Figure 1 B), correlates with e xtensi v e interactions between E 3-1 and both J 1 and V 1, as predicted by Hi-C (Figure 1 C) and confirmed by 3C (Supplementary Figures S1A and S1B).
Two lymphocyte-specific transcription factors, PU.1 ( 38 ) and IRF4 ( 39 ) bind to a composite IRF4 / PU.1 site in E 3-1 ( 39 , 40 ) and have been shown to be important to its function ( 41 ).To further verify this, luciferase constructs were generated with the J 1 promoter ± E 3-1 sequences and electroporated into the transformed pre-B cell line, 103 / BCL-2 ( 24 ).Inclusion of E 3-1 results in ∼3-fold more luciferase activity compared to the J 1 promoter alone whereas single mutations within the core consensus motifs of PU.1 or IRF4 cause a significant decrease or e v en loss of luciferase activity compared to the wild-type enhancer (Supplementary Figure S1C).
To determine if enhanced V 1 and J 1 transcription corr elates with incr eased IRF4 binding at E 3-1, chroma tin immunoprecipita tion (ChIP) was performed.Consistent with previous findings ( 7 ), IRF4 binding at E 3-1 increases ∼3-fold from primary pro-to pre-B cells (Figure 1 D).A small, but r eproducible, incr ease in IRF4 binding is also detected at both V 1 and J 1 promoters in pre-B cells (Figure 1 D); IRF4 does not directly bind to these promoters but instead, the observed increase may result from enhancer-promoter inter actions.By contr ast, ChIP analyses show PU.1 binding at E 3-1 does not change significantly between pro-and pre-B cells (Figure 1 D).PU.1 has a high affinity for its binding motif, whereas IRF4 interacts onl y weakl y with DN A in the absence of PU.1 ( 42 ).From this, and previous studies ( 24 , 39 , 40 ), it is feasible that PU.1 provides a binding platform for IRF4.

Induction of the mouse IG locus enables temporal investigation of coordinate enhancer-mediated activation
Previous data from our lab showed that equipping pro-B cells with elevated, pre-B cell levels of IRF4 is sufficient to completely trigger Ig locus activation ( 7 ).The ability of just a single transcription factor to cause such profound changes at a small, well-defined locus provides a rare opportunity to follow enhancer-mediated locus activation temporally and gain novel insights into key regulatory e v ents.We ther efor e gener ated tr ansgenic mice that express an inducible IRF4, namely IRF4-ER, where the oestrogen receptor hormone binding domain is expressed in frame with IRF4 (Supplementary Figure S1D).Using pro-B cell cultures from these mice, we find that V 1 and J 1 transcription are coordinately and sharply increased between 7 and 8 h of addition of tamoxifen (Supplementary Figure S1D) wher eas r ecombination begins to incr ease shortly ther eafter and continues to increase until 15 hpi (Supplementary Fig- ure S1E), consistent with the r equir ement for non-coding transcription to activate recombination ( 43 ).
To have sufficient cells to investigate this activation in more detail, we next generated a pro-B cell line that also e xpresses inducib le IRF4 (IRF4-ERT2; Figure 2 A), but where a modified oestrogen receptor hormone binding domain was used to reduce non-specific activation ( 22 ).Single cell clones were selected that express IRF4-ERT2 at pre-B cell le v els (Supplementary Figure S1F), resulting in the cell line, PIPER-15 (Figure 2 A).Temporal RT-qPCR analyses show that addition of the oestrogen antagonist, 4hydro xytamo xifen, results in a modest increase in V 1 and J 1 transcription in PIPER-15 cells from 0 to 8 h postinduction (hpi), followed by a sharp increase from 8 to 12 h (Figure 2 B); this correlates well with the changes in primary cells, albeit with slightly delayed kinetics.Consistent with its regulatory role, IRF4-ERT2 translocates to the nucleus, reaching its highest le v el at just 2 h post-induction (Figure 2 C).Furthermore, IRF4-mediated activation is dependent on PU.1: knock-down studies show that loss of PU.1 significantly reduces V 1 and J 1 transcription and IRF4 binding (Supplementary Figures S2A-C).
To investigate the link between activator binding to the enhancer and target gene activation, temporal ChIP analysis was performed.Remar kab ly, IRF4 binding to E 3-1 increases dramatically from 0 to 4 hpi, followed by only a slight increase from 4 to 12 hpi (Figure 2 D), suggesting that enhancer binding by IRF4 is an early e v ent in Ig activation.A limited but clearly detectable increase of IRF4 binding to the V 1 promoter is also observed at 8 hpi (Figure 2 D) but significant enrichment of IRF4 at the J 1 promoter was not detected (Supplementary Figure S2D).Consistent with this, J 1 transcription is substantially r epr essed in PIPER-15 cells compared to primary pre-B cells (Supplementary Figure S2E), e v en though V 1 and J 1 show a similar fold-induction (Figure 2 B).Reduced J 1 transcription may be explained, howe v er, by binding of the transcriptional r epr essor, STAT5, to the J 1 promoter (Supplementary Figure S2F), where STAT5 is likely activated by v-Abl kinase ( 44 ) in the Ab l-kinase-deri v ed cell line, PIPER-15.Consistent with this, J 1 transcription is significantly increased upon ad-dition of imatinib to inhibit Abl-kinase (Supplementary Figure S2G).
Chromatin contraction between an enhancer and its cognate promoter is r equir ed for transcription activation and ther efor e, it would be expected that the interaction frequency between the E 3-1 enhancer and V 1 and J 1 promoters will increase post-induction.Temporal chromatin conformation capture (3C) analysis confirmed that this is indeed the case by 8 hpi, just before enhanced V 1 transcription is observed (Figure 2 E).This is also consistent with the increased enhancer-promoter contacts observed in primary cells (Supplementary Figures S1A and S1B).Together, these data show a coordinate increase in V 1 and J 1 transcription as well as striking temporal changes in their interactions with E 3-1, implying that PIPER-15 cells are a good model to investigate the mechanisms that underpin coordina te, enhancer-media ted activa tion.

IRF4 regulates sequential recruitment of diverse transcription factors to trigger enhancer-promoter interactions
Enhancer-media ted activa tion r equir es the coordinated action of multiple transcription factors, including histone modifying enzymes, lineage-specific transcription factors and ar chitectur e factors ( 5 ).To identify the proteins involved in E 3-1-mediated activation, published ChIP-seq data from primary pro-B cells and pro-B-deri v ed cell lines (pr e-B for YY1) wer e analysed.In addition to IRF4 and PU.1, significant enrichment of E2A, p300, Mediator and YY1 is observed at the E 3-1 enhancer (Figure 3 A).Although activator binding is detected in pro-B cells, this may be explained by low le v els of Ig transcription in these cells, which is increased 8-fold upon transition to pre-B cells ( 7 , 45 ).To determine which factors play key roles in the sharp, coordinate increase in V 1 and J 1 transcription, we capitalised on the inducible nature of PIPER-15 cells to systematicall y anal yse the temporal recruitment of each factor.
The basic helix-loop-helix (bHLH) transcription factor E2A interacts with IRF4 ( 19 ) and knockout studies showed that it is crucial to promote non-coding transcription of unrearranged Ig gene segments in pre-B cells ( 46 ).Consistent with this, ChIP-Seq data show substantial E2A binding at the E 3-1 enhancer (Figure 3 A) and ChIP-qPCR detects a significant increase in E2A at both E 3-1 and V 1p fr om pr o-to pr e-B cells (Supplementary Figur e S3A).Complementary temporal ChIP analyses in PIPER-15 cells suggest that E2A is enriched at E 3-1 prior to induction and that binding increases gradually following IRF4 binding (Figure 3 B).A similar temporal change is observed at the V 1 promoter although here, E2A binding is much lower (Figure 3 B).p300 is a histone acetyltr ansfer ase that exerts its function in concert with numerous transcription factors and media tes acetyla tion of histones close to enhancers and promoters to generate more fle xib le and accessible chroma tin ( 47 ).Co-immunoprecipita tion experiments demonstra ted tha t E2A directly interacts with se v eral histone acetyltr ansfer ases, including p300, that act in synergy with p300 to activate the Ig locus ( 48 , 49 ).Similar to E2A, there is a peak of p300 binding at E 3-1 (Figure 3 A) in primary pro-B cells, which is significantly increased in pre-B cells (Supplementary Figure S3A).Temporal ChIP analysis in PIPER-15 cells shows that the largest increase of p300 binding at E 3-1 is from 0 to 4 hpi, followed by a more gradual increase to 12 hpi (Figure 3 C).A moderate, but reproducible, increase of binding is also observed at the V 1 promoter (Figure 3 C).Consistent with an increase in p300 activity at the enhancer and promoter, ATAC-seq and H3K27ac ChIP-seq data show increased chromatin acetyla tion / accessibility a t E 3-1 and its target promoters fr om primary pr o-and pre-B cells (Supplementary Figure S3B).These changes in E2A and p300 binding ther efor e likely contribute to V 1 acti vation.Howe v er, neither shows the binding kinetics consistent with the sharp increase in V 1 transcription from 8-12 hpi and neither is known to stabilise enhancer-promoter interactions.
The Mediator comple x, howe v er, is an e volutionarily conserved, multi-subunit protein complex that plays an essential role in enhancer-promoter communications ( 50 ).This complex consists of more than 30 subunits which are organized into four distinct modules: the head, middle, tail and kinase modules ( 51 ).The head and middle modules interact with RNAPII and other components of the preinitiation complex ( 52 , 53 ) whereas tail module subunits physically interact with enhancer-bound transcription activators ( 54 ).Thus, it was suggested tha t Media tor provides a physical bridge between transcription activators at enhancers and the preinitiation complex at promoters ( 50 ), a role supported by recent short term knock-down studies and high resolution analysis of long range interactions ( 55 ).Pr evious co-immunopr ecipitation assays r e v ealed that IRF4 directly interacts with MED23 ( 56 ), which is the largest subunit in the tail module and is essential for early B cell de v elopment ( 57 ).To determine if MED23 is required for Ig activa tion, shRNA-media ted knockdown was performed, resulting in a dramatic reduction in MED23 protein le v els in cells e xpressing shRNA against MED23 (shMED23) compared to scrambled shRNA (shSCR; Supplementary Figure S3C, left).Crucially, V 1 transcription is also significantly decreased in shMED23 PIPER-15 cells (Supplementary Figure S3C) as is the interaction between E 3-1 and the V 1 and J 1 promoters following induction (Supplementary Figure S3C, right).These data ther efor e suggest that MED23 is essential for the coordinate activation of V 1 and J 1 transcription.
Ideally, the role of MED23 would be further investigated via temporal ChIP but ChIP-grade anti-MED23 antibodies are not available.Such antibodies are available, howe v er, against MED1, the largest subunit of the Mediator complex, located in the middle module.Since this is part of the functional core of Mediator ( 58 ), analysis of MED1 binding is expected to mirror that of MED23.To first verify that MED1 is r equir ed for V 1 transcription, its expression was knocked down: Western blotting confirmed that MED1 protein le v els are dramatically decreased (Supplementary Figure S3D), correlating with significantly reduced V 1 transcription (Supplementary Figure S3D); crucially, howe v er, transcription of Irf4 , Ctcf and Smc1a is not significantly altered in either MED1 or MED23 knockdown cells (Supplementary Figure S3E) suggesting loss of MED1 / 23 does not uniformly decrease transcription.Next, ChIP analysis was used to investigate how Mediator contributes to V 1 activation.MED1 binding to E 3-1 and V 1 increases significantly from pro-to pre-B cells (Supplementary Figure S3A) as well as in PIPER-15 cells following IRF4 induction.Here, the biggest relati v e increase at E 3-1 is from 0 to 4 hpi but further gradual increases are observed to 12 hpi (Figure 3 D).Compared to the enhancer, MED1 binding to the V 1 promoter is low but reproducible and correlates with V 1 transcription, albeit without the sharp increase between 8 and 12 hpi (Figure 3 D).Together, these data suggest that Mediator recruitment by IRF4 is vital for Ig transcription and may contribute to enhancer / promoter bridging.
YY1 is a ubiquitously expressed, zinc finger DNA binding protein that activates or r epr esses transcription, depending on the context in which it binds ( 59 ).YY1 plays an important role in chromatin folding of the IgH locus, where a YY1 conditional knockout led to decreased chromatin looping ( 16 ).Published ChIP-seq data indicate that YY1 is also enriched at the E 3-1 enhancer in pre-B cells (Figure 3 A).To investigate if YY1 influences V 1 transcription and / or Ig chromatin organization, shRNA against YY1 was expressed in PIPER-15 cells.Both YY1 protein le v els (Supplementary Figure S3F) and V 1 transcription were dramatically reduced (Supplementary Figure S3F), as is the E 3-1 interaction frequency with V 1 (and J 1) follow-ing induction, as determined by 3C (Supplementary Figure S3F).Control experiments confirmed that transcription of Irf4 , Ctcf and Smc1a are relati v ely unchanged (Supplementary Figure S3G).We find further that YY1 binding to E 3-1 and V 1 is significantly increased from pro-to pre-B cells (Supplementary Figure S3A).These data ther efor e imply that YY1 is essential for V 1 transcription activation and chromatin organization of the Ig locus.To determine at which stage YY1 is r equir ed, temporal ChIP analysis was carried out following IRF4 induction.Intriguingly, YY1 is enriched at both the E 3-1 enhancer and V 1 promoter but its binding only increases significantly from 8 to 12 hpi at both r egions (Figur e 3 E).This ther efor e corr elates very well with the sharp increase in V 1 and J 1 transcription and suggests that YY1 is pivotal to this increase.
To better understand how RNAPII is recruited to achie v e this increased transcription, ChIP experiments with antibodies against serine-5 and serine-2 phosphorylated Cterminal domain were carried out.RNAPII is already present at the E 3-1 enhancer and V 1 promoter at low le v els prior to IRF4 induction, consistent with low le v els of V 1 transcription in pro-B cells ( 7 ).Upon induction, both serine-5 and serine-2 phosphorylated RN APII graduall y increase at the E 3-1 enhancer (Supplementary Figures S4A  and S4B), concomitant with enhancer activation and correlating with Mediator binding to the enhancer (Figure 3 D).At the V 1 promoter, an increase in serine-5 phosphorylated RNAPII is observed at 8 hpi, corresponding to increased 3C interactions between E 3-1 and V 1 (Figure 2 E).Binding of serine-5 phosphorylated RNAPII then decreases concomitant with a significant increase of promoterbound serine-2 phosphorylated RNAPII between 8 and 12 hpi.These data ther efor e suggest that RNAPII is initially recruited to the enhancer and transferred to the promoter via enhancer-promoter interactions during transcriptional activation.It is also notable that the sharp increase in serine-2 phosphoryla ted RNAPII a t the V 1 promoter correlates with increased YY1 binding and V 1 transcription.

IRF4 mediated formation of an enhancer hub is essential for Ig activation
Whilst the above analyses identify which activators play important roles in V 1 activation, they do not explain how J 1 is coordinately upregulated, nor do they show if other gene regulatory elements are required.Antigen receptor loci typicall y contain m ultiple gene segments and corresponding regulatory DNA elements that can span mega-base sized chromatin regions.To characterize additional cis-acting elements, published ATAC-seq and ChIP-seq data from pro-B cells were reprocessed and mapped to the murine Ig locus (Figure 3 A).In addition to E 3-1, four further regions of open chromatin were found in the 3 domain of Ig , namely HSCV 1, HSC 1, HSE-1 and HSE-2 in primary pro-B cells (Figure 3 A) and PIPER-15 cells (Supplementary Figure S4C).
Intriguingly, two of these sites, HSV 1 and HSE-2, lie at the very 5 and 3 of the 3 domain and show peaks of CTCF and cohesin (RAD21) binding.These essential architecture factors generate chromatin loops that separate the genome into di v erse domains and thus may create an insulated neighbourhood domain in the 3 region of the Ig locus (Figure 4 A).Consistent with this idea, CTCF typically media tes chroma tin loops between convergent CTCF motifs ( 60 ), which is the orientation observed at HSV 1 and HSE-2 (Figure 4 A).Not only this, but Hi-C data from pre-B cells indicate substantial interactions between HSV 1 and HSE-2 (Figure 1 C) and ChIP-qPCR experiments show a marked enrichment of CTCF and SMC1A (cohesin subunit) at HSV 1 and HSE-2 that is unaltered in PIPER-15 cells following IRF4 induction (Figure 4 A) and between primary pro-and pre-B cells (Supplementary Figure S4D).These data ther efor e imply that CT CF / cohesin connects HSV 1 and HSE-2 via a chromatin loop to create the Ig locus 3 domain.Notably, this loop brings HSE-1 and the V 1 promoter into closer proximity, which may facilitate V 1 activation (Figure 4 A, lower).
We next examined which elements might cooperate within the large CTCF / cohesin-generated loop to orchestrate Ig locus activation.Similar to E 3-1, both HSC 1 and HSE-1 are open chromatin regions with a high le v el of H3K27ac and p300 binding (Figure 3 A) and thus display the characteristics of acti v e enhancers.Consistent with this idea, ChIP-seq data from pro-B cells shows transcription factor binding peaks at HSE-1 and HSC 1 that are very similar to those at E 3-1 (Figure 3 A).Moreover, the IRF4 ChIP-qPCR signal is highly enriched at E 3-1 and HSE-1 (Figure 4 B), possibly due to recruitment by pre-bound PU.1 ( 39 , 42 ) whereas low le v els of IRF4 are present at HSC 1 where PU.1 is absent.These data therefore imply that the newly identified enhancer-like elements HSE-1 and HSC 1 play an integral role in Ig locus activation.In support of this, significant interactions among these three enhancer elements are seen in Hi-C data (Figure 1 C), suggesting that they may form an enhancer hub.
To test this idea, temporal 3C analysis was performed using E 3-1 as a viewpoint.Prior to Ig activation, E 3-1 exhibits limited contacts with HSE-1 and HSC 1 or with the unrearranged V 1, J 1 and J 3 gene segments.Following induction, chromatin contacts do not change dramatically by 4 hpi.Remar kab ly, howe v er, a substantial increase in interaction frequency between E 3-1 and V 1, J 1 and J 3 occurs by 8 hpi, with a further increase by 12 hpi (Figure 4 C), mirroring significant increases in transcription (Figure 2 B).Not only this, but the interaction frequency between E 3-1 and HSE-1 as well as HSC 1 correlates well with the changes in interactions between E 3-1 and V 1, J 1 and J 3. These data ther efor e suggest that E 3-1 interacts with HSC 1 and HSE-1 to from an enhancer hub and that the target genes, V 1, J 1 and J 3, ar e concurr ently brought into proximity of this hub, allowing their coordinate activation.
To further investigate the enhancer hub idea, we next separately knocked out the PU.1 / IRF4 binding sites within the E 3-1 and HSE-1 enhancers using CRISPR / Cas9 (Supplementary Figure S4E and Supplementary Table 2).Changes in V 1 and J 1 transcription were then determined as well as alterations in enhancer-promoter interactions from the HSE-1 and E 3-1 viewpoints.Consistent with idea that IRF4 is central to locus activ ation, remov al of its motif from either enhancer results in a significant reduction in both V 1 and J 1 transcription (Supplementary Figure S4F), that correlates with a dramatic loss of both enhancerenhancer and enhancer-promoter interactions throughout the entire locus (Supplementary Figures S4G and S4H).The fact that loss of IRF4 binding to just one enhancer, either E 3-1 or HSE.1, causes such fundamental changes to the whole locus, supports the idea of coordinated enhancer hub formation.
The striking similarity of transcription factor motifs at E 3-1, HSC 1 and HSE-1 suggests that they may share comparable dynamic transcription factor binding profiles tha t could facilita te enhancer hub forma tion.To investiga te this, ChIP analyses of IRF4, E2A, p300, MED1 and YY1 were performed at E 3-1, HSE-1 and HSC 1 in pro-B, pre-B and PIPER-15 cells.Temporal analyses showed that, similar to its recruitment to E 3-1, IRF4 binding to HSE-1 is an early e v ent that reaches its maximal le v el at 4 hpi, (Figure 4 B).IRF4 binding to HSC 1 shows a similar temporal pattern of recruitment, although here, in the absence of PU.1 (Figure 3 A), binding occurs at only low le v els (Figure 4 B).Just as for E 3-1, binding to HSE-1 and HSC 1 is also significantly increased from pro-B to pre-B cells (Supplementary Figure S5).
E2A and p300 binding to HSE-1 and HSC 1 also show a similar temporal pattern of recruitment to that seen at E 3-1, with significantly increased binding at 8 and 12 hpi (Figure 4 B), which is also consistent with data fr om pr o-and pr e-B cells (Supplementary Figur e S5).Together, these data suggest that IRF4 interacts directly with E 3-1, HSE-1 and HSC 1 and this increased IRF4 binding results in recruitment of E2A and p300 to generate open chromatin regions.
Formation of the enhancer hub r equir es the constituent enhancers to be brought into closer proximity.To determine if Mediator is involv ed, pub lished MED1 ChIP-seq fr om pr o-B cells was analysed.As can be seen in Figure 3 A, MED1 is already present at HSE-1 and HSC 1 at low levels; like wise, low le v els of IRF4 are found at both elements consistent with low le v el locus acti vity in pro-B cells and it is possible Mediator is recruited through direct interactions with IRF4.Following induction of PIPER-15 cells, a gradual increase in MED1 binding to HSE-1 and HSC 1 is observed (Figure 4 B), mirroring its binding to E 3-1 (Figure 3 D), and consistent with the significantly increased binding between pro-and pre-B cells (Supplementary Figure S5).To determine if Mediator is essential to establish interactions that lead to enhancer hub formation, 3C analysis was performed in MED23 knock-down PIPER-15 cells, with and without IRF4 induction: E 3-1, HSE-1 and HSC 1 (Figure 5 A) contacts are dramatically decreased, as are interactions between E 3-1 and the J 1, V 1 and J 3, gene segments (Figure 5 A).These data therefore imply that Mediator is vital for IRF4-mediated formation of the Ig enhancer hub and for interactions with gene segment promoters, leading to their coordinate activation.
It is notable that knockdown of MED1 (and YY1) reduces the interactions seen at 0 hpi compared to those seen with the scrambled RNA (orange with black plots, respecti v ely).This further correlates with reduced V 1 transcription at 0 hpi in the presence of shMED1 / 23 versus shSCR (and shYY1 versus shSCR; Supplementary    The height of curves between E 3-1 and other genomic fragments r epr esents the average value of interaction frequency obtained from three experimental repeats (Supplementary Table 1).Data were normalized using an interaction within the Ercc3 locus.The plots to the right show the significance of the difference in interactions at 12 hpi between shSCR and shMED23.( B ) Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 viewpoint in PIPER-15 cells expressing an shRNA targeting Yy1 .The height of curves between E 3-1 and other genomic fragments r epr esents the average value of interaction frequency obtained from three experimental repeats (Supplementary Table 1).Data were normalized using an interaction within the Ercc3 locus.The plots to the right show the significance of the difference in interactions at 12 hpi between shSCR and shYY1.Notably, the biggest increase in locus interactions is between 4 and 8 hpi (Figur e 4 C) wher eas the biggest increase in YY1 binding is between 8 and 12 hpi.Howe v er, considerab le YY1 binding is observed prior to locus induction; this could stabilise long r ange inter actions as they are f ormed and ma y explain w hy knockdown of YY1 impacts so significantl y on locus folding.
Figure S3C, D, F).This may be because knock-down of these factors causes loss of the low le v el (IRF4-dependent) activity of the Ig locus in pro-B cells ( 7 ).
To measure the impact of YY1 on enhancer hub formation, temporal analysis of YY1 binding was performed in PIPER-15 cells.As can be seen in Figure 4 B, YY1 occupancy at HSE-1 and HSC 1 dramatically increases from 8 hpi to 12 hpi, mirroring its binding to E 3-1 (Figure 3 E) and the large increase in V 1 and J 1 transcription following induction (Figure 2 B).Significantly increased YY1 enrichment at these two enhancers is also observed in pre-B compared to pro-B cells (Supplementary Figure S5) and notably, the fold-change in binding from pro-to pre-B cells as well as from 8 to 12 hpi in PIPER-15 cells is very similar at all enhancer-like elements, including E 3-1.To determine if YY1 binding also modulates locus folding, its expression was depleted in PIPER-15 cells (Supplementary Figure S3F).Remar kab ly, this resulted in a significant reduction of the enhanced interactions at 12 hpi between E 3-1 and HSC 1 (Figure 5 B and Supplementary Figure S3F) as well as between E 3-1 and its target genes V 1, J 1 and J 3 (Figure 5 B), correlating with diminished V 1 transcription (Supplementary Figure S3F).Furthermore, knockout of the YY1 site in HSC 1 almost eliminated E 3-1 / HSC 1 interactions as well as E 3-1 interactions with V 1, J 1 and J 1, and resulted in significantly reduced V 1 transcription ( 61 ).YY1 ther efor e appears to be pivotal to the interactions between the enhancers and target genes and the coordinate activation of otherwise distant V 1, J 1 and J 3 promoters, although not E 3-1 / HSE.1 inter actions.Tempor al analyses suggest that YY1 functions later than Mediator, perhaps by stabilising pre-formed interactions; nonetheless, the dramatic disruption of locus folding in the absence of YY1, implies that its function is vital.

Antisense eRNAs encoded by E 3-1 r epr ess YY1 recruitment
YY1 has a relati v ely low affinity for DNA ( 62 ) and although its binding may be stabilised via IRF4-interacting proteins, such as p300 ( 63 ), exactly how YY1 binding is stabilised, is unclear.Gi v en its vital role in locus folding, the mechanism of YY1 stabilisation could be pivotal to locus activation.Enhancer RNAs (eRNAs) are a sub-class of noncoding RNAs that are transcribed from acti v e enhancers and have been demonstrated to be involved in enhancerpromoter loop formation and target gene activation ( 64 ).Pre vious pub lica tions demonstra ted tha t eRNAs can interact with di v erse transcription factors, including cohesin ( 65 ), Mediator ( 66 ), YY1 ( 67 ) and p300 ( 68 ).Notably, RN A-seq reads ma p to E 3-1 (Supplementary Figure S6A) and E 3-1 eRNA expression levels increase significantly from primary pro-B to pre-B cells (Supplementary Figure S6B), suggesting that they may be important to Ig locus activation.
To further investigate this, eRNA expression was analysed temporally via RT-qPCR following IRF4 induction.
As can be seen in Figure 6 A, total E 3-1 eRNA le v els show a marginal increase between 4 and 8 hpi, prior to increased YY1 binding at E 3-1 (Figure 3 E).A large increase is observ ed howe v er, between 8 and 12 h which correlates well with the largest increase in YY1 binding (Figure 3 E).YY1 was previously demonstrated to be trapped by RNAs tethered at enhancers ( 67 ) and these data suggest that increased YY1 binding at E 3-1 may be facilitated by eRNAs.
Similar to mRN As, eRN As are also transcribed by the RNAPII machinery.Howe v er, their 3 ends are processed by the Integrator complex which facilitates eRNA maturation and their release from transcribing RNAPII ( 69 ).Consistent with a role for eRNA, temporal ChIP analysis of Integrator (the INTS11 subunit) binding in PIPER-15 cells shows that it reaches its highest le v el at E 3-1, HSE-1 and HSC 1 at 4 hpi (Figures 4 B and 6 B), just prior to increased eRNA le v els.Increases in Integrator occupancy are also observed at all three enhancer-like elements and the V 1 promoter fr om pr o-to pr e-B cells (Supplementary Figur es S5 and S6C).
Genome-wide analysis suggests that the majority of enhancers are transcribed bidirectionally ( 70 ) and GRO-seq data fr om pr o-B cells identifies a number of reads that map to both sense and antisense strands of the E 3-1 enhancer as well as HSC 1 and HSE-1 (Figures 6 C and S6D).Temporal analysis of sense and antisense E 3-1 eRNA expression, following re v erse tr anscription with str and-specific primers, shows that sense E 3-1 eRNA starts to increase between 4 and 8 hpi (Figure 6 D), whereas changes in antisense eRNA are much lower than sense eRNA.Consequently, the relati v e amount of sense eRNA increases compared to antisense.
Notab ly, the le v el of V 1 transcription is altered in the presence of sense or antisense shRNA at 0 hpi, in a similar way to at 12 hpi, when compared to the respecti v e scrambled control.This may be because the Ig locus is already acti v e at low le v els at 0 hpi ( 7 ) and knockdown of the eRN As likel y affects both the basal (pro-B-like), as well as induced V 1 transcription.Consequently, the foldinduction (comparing 0 to 12 hpi) appears similar between the scrambled shRNA control and the sense or antisense eRNA knockdown.Gi v en that the knockdown affects both uninduced and induced transcription, we compare scrambled and knock-down shRNA levels at either 0 or 12 hpi (Supplementary Figures S6E and S6F).
Pre vious pub lica tions showed tha t eRNAs are essential to establish enhancer-promoter interactions ( 65 , 72 ).Theref ore, to in vestigate the impact of sense and antisense E 3-1 eRNAs on formation of the V 1-E 3-1 chromatin loop, 3C analysis of E 3-1 to V 1 interactions was performed in shsE 3-1e and shasE 3-1e PIPER-15 cells, respecti v ely.Consistent with the observed transcription changes, V 1-E 3-1 interactions are significantly reduced in shsE 3-1e PIPER-15 cells at 12 hpi (Figure 6 E and Supplementary Figure S6H), indicating that sense E 3-1 eRNA is vital to establish enhancer-promoter chromatin loops.Howe v er, remar kab ly, E 3-1 to V 1 interaction frequency is significantly increased in shasE 3-1e PIPER-15 cells (Figure 6 E, lower and Supplementary Figure S6I), implying that antisense eRNAs r epr ess enhancer / promoter loop formation and reduce target gene transcription.
Enhancer RNAs are known to exert their functions by interacting with di v erse transcription factors (66)(67)(68).Temporal ChIP analysis showed that E 3-1 eRNAs increase just prior to increased YY1 enrichment at E 3-1, suggesting that expression of eRNAs may be a pr er equisite for stable YY1 binding; conversely, diminished locus folding in eRNA knock-down cells may be caused by reduced YY1 binding.To test this idea, ChIP-qPCR analysis of YY1 binding to E 3-1 was performed in the shsE 3-1e and shasE 3-1e PIPER-15 cells.As can be seen in Figure 6 F, knockdown of the sense E 3-1 eRNA leads to decreased YY1 occupancy at E 3-1, suggesting that eRNA-mediated chromatin folding is indeed associated with YY1 binding.Intriguingly, howe v er, YY1 binding to E 3-1 is increased in the antisense E 3-1 eRNA knock-down cells (Figure 6 F), indicating that the antisense eRNAs r epr ess YY1 recruitment.
The bidirectional sense and antisense E 3-1 eRNAs appear to arise from different regions of the E 3-1 enhancer (Figure 6 C) but short regions of homology (Supplementary Figur e S6J) ar e pr esent.It ther efor e seemed possible that antisense eRNAs interact with sense eRNAs to regulate YY1 recruitment to enhancers.To test this, sense and antisense E 3-1 eRNA were generated via in vitro transcription and RN A-RN A hybridization experiments performed.Interactions between antisense and sense E 3-1 eRNAs are indeed observ ed, e videnced by duple x forma tion in vitr o (Figure 6 G); by contrast, control experiments using a random RNA that lacks sequence homology, failed to hybridise (Supplementary Figure S6K).To further investigate the role of antisense enhancer in vivo , RN A imm unoprecipitation was performed; this showed that YY1 pulls down ∼2.5fold more sense eRNAs in the antisense eRNA knock-down cells compared to PIPER-15 cells (Figure 6 H).Together, these data indicate that antisense eRNAs interact with sense eRNAs to suppress YY1 recruitment and stable locus activation.

DISCUSSION
Enhancer-media ted activa tion is vital for the correct le v els of transcription at the right de v elopmental stages.Whilst numerous studies have shown that enhancers trigger activ ation b y physically interacting with their cognate promoters, antigen receptor loci pose a unique problem in that non-coding transcription must be coordinately upregulated through at least two distant, complementary gene segments prior to their r ecombination.Her e, we capitalised on the finding that increased le v els of just a single transcription factor, IRF4, is sufficient to completely activate the murine Ig locus, to de v elop a system in which Ig gene transcription can be reliably induced.Using this novel inducible system, we confirm for the first time, that non-coding transcription through V and J gene segments is indeed coordinately upregulated.We then systematicall y anal ysed the temporal recruitment of transcription activators, as well as long range chromatin folding and find that three enhancer elements in the 3 domain of the Ig locus show remar kab ly similar dynamics of activator binding.Our temporal analyses show further that these enhancers are brought together into an activating hub, concomitant with the recruitment of V and J promoters and their transcription activation.Colocalisation of enhancers and promoters within the same activating hub is thus central to coordinate V 1 and J 1 activation.Gi v en that other antigen receptor loci also undergo locus folding, it is highly feasible that similar mechanisms are adopted to coordinately activate their complementary gene segments, prior to recombination (73)(74)(75).
Our studies show further that establishment of the functional enhancer-promoter hub r equir es di v erse transcription factors, including general transcription factors, lineage-specific transcription factors, histone modifiers, architecture transcription factors as well as eRNAs.By following transcription activation temporally, we could deduce which e v ents correlate most closely with both locus folding and transcription upregulation and thus are potentially key r egulatory steps.A sharp incr ease in V 1 and J 1 transcription is observed between 8 and 12 hpi that correlates extremely well with YY1 binding to all three enhancers, suggesting that this is a key e v ent.YY1 functions as a transcription factor, but can also help to establish chromatin loops, especially enhancer-promoter loops ( 76 ).Consistent with the latter role, knockdown of YY1 reduces V 1 noncoding transcription and results in a se v ere disruption of long range chromatin interactions, without a significant loss of transcription at other loci.These data ther efor e suggest that YY1 binding is important to stabilise the acti v e chromatin hub and to thereby achie v e consistent transcription activation.
Our knock-down studies demonstrate that, in addition to YY1, Mediator is also vital for V 1 transcription and Ig locus folding.Remar kab ly, long r ange inter actions are altered very similarly upon loss of either transcription factor, suggesting that they interact with very similar regions, but function independentl y.Notabl y, enhancer-promoter, as well as enhancer-enhancer interactions are disrupted, which supports the idea that an acti v e hub is formed between enhancer-and promoter-bound transcription factors.Gi v en that there is an increase in long range interactions between 4 and 8 hpi without a change in YY1 binding, it appears tha t Media tor may establish enhancer / promoter interactions that are then stabilised by YY1.
The bridging role of Mediator between enhancers and promoters, although debated, has recently been shown to be important at 20 loci regulated by super-enhancers ( 55 ), and can explain how Mediator establishes the long range interactions ( 50 ).YY1, howe v er, has both DNA and RNA bind-ing activity as well as an intrinsically disordered domain that is distinct from its DNA binding domain ( 77 ).Consensus YY1 DNA binding motifs are not present at the V 1 nor J 1 / 3 promoters and whilst we find eRN A likel y stabilises YY1 binding at E 3-1, its recruitment to promoters is unclear.Gi v en the formation of an acti v e hub, YY1 localisation to promoters may involve interactions between intrinsically disordered domains in both YY1 and other transcription activators in the enhancer / promoter hub ( 78 ).
Notably, knockout of either E 3-1 or HSE-1 enhancer results in a substantial reduction in V 1 and J 1 transcription and a significant reduction in most long range interactions.The r equir ement for both enhancers to achie v e transcription activation may be explained if activation depends on threshold le v els of acti vation potential within the enhancer-promoter hub.Although temporal analyses show that some transcription activation occurs prior to complete binding of all transcription factors, transcription levels are only modest and full transcription activation is only achie v ed upon increased YY1 binding, consistent with a vital role for YY1 in stabilising long range interactions.
Although YY1 interacts directly with Mediator ( 79 ) and p300 ( 63 ), these proteins bind to E 3-1 as early e v ents in Ig locus activation; YY1 binding, however, increases much la ter, suggesting tha t YY1 is recruited by other factors.YY1 contains RNA binding domains and previous studies demonstrated that YY1 can be recruited via enhancertethered eRNAs ( 67 ).Temporal analysis of the expression of eRNAs encoded by E 3-1 shows that changes in the le v el of eRNAs correlate very well with changes in YY1 binding: eRNAs start to increase from 4 hpi, just prior to increased YY1 binding, with their largest increase between 8 and 12 hpi, concomitant with the sharp increase in YY1 binding.It is also notable that serine-2 phosphorylated RNAPII at the V 1 promoter increases significantly between 8 and 12 hpi, correlating with the sharp increases in V 1 transcription, YY1 binding and sense eRNA.By contrast, binding of serine-5 phosphorylated RNAPII decreases from 8 to 12 hpi.Pre vious studies hav e demonstrated eRNA can recruit CDK9 of the P-TEFb complex ( 80 ) and it ther efor e seems possible that sense eRNA fulfils a second function of recruiting the P-TEFb complex to activate transcription.
Integrator is r equir ed for eRNA biosynthesis and its binding to E 3-1 reaches its highest le v el at 4 hpi, just prior to the increase of eRNAs.Consistent with concerted acti vity between acti va tors, Integra tor directly interacts with Mediator ( 81 ); gi v en that Mediator binding to E 3-1 is an early e v ent, it is feasib le tha t Media tor recruits Integra tor to modulate changes in eRNAs.
Most enhancers are transcribed bidirectionally but the function of the non-dominant eRN A is largel y unknown.Our studies show that remar kab ly, the less dominant, antisense E 3-1 eRNA r epr esses YY1 recruitment, raising the question of just how this is achie v ed.Tertiary structure is essential for eRNAs to recognize their binding partners.For example, 40% of sense eRNAs possess a functional eRNA regulatory motif (FERM) which can mediate interactions with di v erse transcription acti vators, including YY1 ( 82 ).Although a FERM is not present in sense E 3-1 eRNA, other such motifs with specific tertiary structures may be present and antisense eRNA might disrupt such structures to block interactions with YY1.Alternati v el y, antisense eRN A might pre v ent other proteins such as p300, MED1, MED12 that interact with eRNAs (83)(84)(85), from tethering sense eRNA to enhancer elements.Consistent with the idea that antisense eRNA interacts with sense eRNA to regulate YY1 binding and / or eRNA tethering, in vitr o RNA hybridiza tion demonstra ted sense / antisense eRNA interactions, despite each eRNA being encoded by distinct sequences.Consequently, the interaction dynamics between sense and antisense eRNAs may regulate transcription factor recruitment.Gi v en that similar sense and antisense eRNAs are present at the other Ig enhancers, it is possible that YY1 is recruited at tethered in similar way at HSE-1 and HSC 1.Consistent with this idea, loss of one enhancer and its corresponding eRNA results in complete loss of Ig chromatin folding.
Together our data show that Ig locus activation r equir es the establishment of the correct chr omatin envir onment that culminates in enhancers and promoters being brought into close proximity, triggering coordinated V and J activation.We demonstrated that HSE-2 and HSV 1 establish a chroma tin loop tha t seals the 3 end of the Ig locus and results in locus contraction, shortening the distance between E 3-1 and the unrearranged gene segments.CTCF / cohesin mediated folding, howe v er, is unlikely to be sufficient to establish enhancer-promoter inter actions.Instead, tempor al 3C analysis showed that two other enhancer-like sequences in addition to E 3-1, namely HSE-1 and HSC 1, have similar transcription factor binding dynamics and interact to form an enhancer hub.From this, we propose a three-step model to explain the chromatin structure changes during Ig locus activation (Graphical abstract): Step 1: Formation of the CTCF / cohesin mediated chromatin loop between HSE-2 and HSV 1; Step 2: IRF4 facilitates locus contraction through recruiting histone modifiers, Mediator and Integrator; Step 3: Upregulation of sense eRNAs causing recruitment of YY1 and stabilisation of Ig folding.As a result of these chromatin changes, the unrearranged gene segments are brought into close proximity of the enhancer hub, establishing, in principle, how their coor dinate acti vation is achie v ed.

DA T A A V AILABILITY
The data underlying this article are available in the article and in its online supplementary material.The ATAC-seq data underlying this article are available under GEO accession GSE239925 and via the UCSC link: http://genome-euro.ucsc.edu/s/asmith151/2023%2D07%2D24%2DPIPER15%2D12hr%2D AT AC%2Dseq .

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.

Figure 1 .
Figure 1.The enhancer, E 3-1, activates target gene transcription via PU.1 and IRF4.( A ) Schematic of the murine Ig locus; the potential gene duplication is indicated by the dashed line.Constant exons (C) are depicted by green rectangles; V gene segments by cyan rectangles; J gene segments by blue rectangles and enhancers by orange ovals.70% of recombination occurs between the V 1 and J 1 gene segments.( B ) Transcription le v els of V 1 and J 1 in primary mouse pro-B and pre-B cells, determined by qPCR.Data are normalized to Hprt expression levels.( C ) Schematic of significant Hi-C interactions in the 3 half of the murine Ig locus.CTCF, RAD21, H3K27ac ChIP-seq, Hi-C and ATAC-seq from pre-B cells were analysed using the HOMER software package and visualised using Circos (RAD21 data from pro-B cells).Significant interactions in 10 kb windows are shown.( D ) IRF4 and PU.1 binding was analysed by ChIP-qPCR in primary mouse pro-B and pre-B cells.The fold enrichment over input at E 3-1, V 1, J 1 and Intgene III (negati v e control region) is shown.All values are normalized to binding at the Intgene III negati v e contr ol.Err or bars show standard error of the mean (SEM) from three biological replicates.

Figure 2 .
Figure 2. De v elopment of an inducib le system to inv estiga te enhancer-promoter interactions.( A ) Schema tic of the genera tion of the pro-B cell line, PIPER-15, that expresses the inducible transgene, Irf4-ERT2 .Bone marrow was extracted from six-week-old mice and immediately infected with the Abelson murine leukaemia virus (A-MuLV) for immortalization.Single cells were isolated by flow cytometry using pro-B specific markers, CD19 and CD43.Retroviruses were generated by transfecting the construct, MSCV-IRF4-ERT2-IRES-GFP, into Phoenix cells, followed by transduction of immortalized pro-B cells by spin-fection.Single, transduced pro-B cells with the highest expression of GFP were isolated by flow cytometry.( B ) The le v el of V 1 and J 1 non-coding transcription was analysed by RT-qPCR in PIPER-15 cells following induction of IRF4-ER with 4-hydro xytamo xifen.A sharp increase is observed from 8 to 12 hpi.Data are normalized to Hprt expression levels.( C ) Analysis of IRF4-ERT2 by western blotting in nuclear extracts of PIPER-15 cells following induction with 4-hydroxytamoxifen.Histone H2A le v els are used as a loading control.( D ) IRF4 binding to the E 3-1 enhancer and V 1 promoter in PIPER-15 cells following induction.The fold enrichment over input at E 3-1, V 1 and Intgene III (negati v e control region) is shown.All values are normalized to binding at the Intgene III negati v e control.( E ) The interaction between E 3-1 and V 1 as well as J 1 was analysed by 3C-qPCR in PIPER-15 cells following induction.Data were normalized using an interaction within the Ercc3 locus.Error bars show standard error of the mean (SEM) from three biological replicates.

Figure 3 .
Figure 3. IRF4 triggers sequential recruitment of di v erse transcription factors.( A ) ATAC-seq and ChIP-seq data of ar chitectur e factors (CT CF, RAD21 and YY1), the enhancer mark (H3K27Ac) and transcription activators (PU.1, IRF4, E2A and MED1) mapped to the 3 half of the Ig locus.All data are fr om pr o-B cells except YY1, which is fr om pre-B cells.( B-E ) E2A, p300, MED1 and YY1 binding at E 3-1 and V 1p were analysed by ChIP-qPCR in PIPER-15 cells following induction.The fold enrichment at E 3-1, V 1p and Intgene III (negati v e control region) is shown.All values are normalized to binding at Intgene III as a negati v e control.Error bars show standard error of the mean (SEM) from three biological replicates.

Figure 4 .
Figure 4. IRF4-media ted forma tion of an enhancer hub activates the Ig locus.( A ) Left, upper: CTCF motifs at HSE-2 and HSV 1 lie in a convergent orientation (green arro ws).Left, lo wer: Schematic sho wing forma tion of the chroma tin loop tha t brings HSE-1 and the V 1 promoter into close proximity.Right: CTCF and SMC1A (cohesin subunit) binding to HSE-2 and HSV 1 measured by ChIP-qPCR in PIPER-15 cells.The fold enrichment at HSE-2, HSV 1 and Intgene III (negati v e control region) is shown.( B ) Binding of IRF4, E2A, p300, MED1, YY1 and INTS11 to HSE-1 and HSC 1 analysed by ChIP-qPCR in PIPER-15 cells following induction.The fold enrichment at HSE-1 and HSC 1 is shown.All values are normalized to binding at Intgene III as a negati v e control.( C ) Temporal 3C analysis of chromatin interactions in the 3 half of the Ig locus.Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 viewpoint in PIPER-15 cells at 0, 4, 8 and 12 hpi.Data were normalized using an interaction within the Ercc3 locus and are the average of three experimental repeats (Supplementary Table1).The plots to the right of the same data show the significance of the difference in interactions between 0 and 12 hpi.Error bars show standard error of the mean (SEM) from three biological replicates.

1
Figure 4. IRF4-media ted forma tion of an enhancer hub activates the Ig locus.( A ) Left, upper: CTCF motifs at HSE-2 and HSV 1 lie in a convergent orientation (green arro ws).Left, lo wer: Schematic sho wing forma tion of the chroma tin loop tha t brings HSE-1 and the V 1 promoter into close proximity.Right: CTCF and SMC1A (cohesin subunit) binding to HSE-2 and HSV 1 measured by ChIP-qPCR in PIPER-15 cells.The fold enrichment at HSE-2, HSV 1 and Intgene III (negati v e control region) is shown.( B ) Binding of IRF4, E2A, p300, MED1, YY1 and INTS11 to HSE-1 and HSC 1 analysed by ChIP-qPCR in PIPER-15 cells following induction.The fold enrichment at HSE-1 and HSC 1 is shown.All values are normalized to binding at Intgene III as a negati v e control.( C ) Temporal 3C analysis of chromatin interactions in the 3 half of the Ig locus.Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 viewpoint in PIPER-15 cells at 0, 4, 8 and 12 hpi.Data were normalized using an interaction within the Ercc3 locus and are the average of three experimental repeats (Supplementary Table1).The plots to the right of the same data show the significance of the difference in interactions between 0 and 12 hpi.Error bars show standard error of the mean (SEM) from three biological replicates.
Figure 4. IRF4-media ted forma tion of an enhancer hub activates the Ig locus.( A ) Left, upper: CTCF motifs at HSE-2 and HSV 1 lie in a convergent orientation (green arro ws).Left, lo wer: Schematic sho wing forma tion of the chroma tin loop tha t brings HSE-1 and the V 1 promoter into close proximity.Right: CTCF and SMC1A (cohesin subunit) binding to HSE-2 and HSV 1 measured by ChIP-qPCR in PIPER-15 cells.The fold enrichment at HSE-2, HSV 1 and Intgene III (negati v e control region) is shown.( B ) Binding of IRF4, E2A, p300, MED1, YY1 and INTS11 to HSE-1 and HSC 1 analysed by ChIP-qPCR in PIPER-15 cells following induction.The fold enrichment at HSE-1 and HSC 1 is shown.All values are normalized to binding at Intgene III as a negati v e control.( C ) Temporal 3C analysis of chromatin interactions in the 3 half of the Ig locus.Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 viewpoint in PIPER-15 cells at 0, 4, 8 and 12 hpi.Data were normalized using an interaction within the Ercc3 locus and are the average of three experimental repeats (Supplementary Table1).The plots to the right of the same data show the significance of the difference in interactions between 0 and 12 hpi.Error bars show standard error of the mean (SEM) from three biological replicates.

Figure 5 .
Figure 5. MED23 and YY1 are essential for Ig locus contraction.( A ) Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 viewpoint in PIPER-15 cells expressing an shRNA targeting Med23 .The height of curves between E 3-1 and other genomic fragments r epr esents the average value of interaction frequency obtained from three experimental repeats (Supplementary Table1).Data were normalized using an interaction within the Ercc3 locus.The plots to the right show the significance of the difference in interactions at 12 hpi between shSCR and shMED23.( B ) Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 viewpoint in PIPER-15 cells expressing an shRNA targeting Yy1 .The height of curves between E 3-1 and other genomic fragments r epr esents the average value of interaction frequency obtained from three experimental repeats (Supplementary Table1).Data were normalized using an interaction within the Ercc3 locus.The plots to the right show the significance of the difference in interactions at 12 hpi between shSCR and shYY1.Notably, the biggest increase in locus interactions is between 4 and 8 hpi (Figur e 4 C) wher eas the biggest increase in YY1 binding is between 8 and 12 hpi.Howe v er, considerab le YY1 binding is observed prior to locus induction; this could stabilise long r ange inter actions as they are f ormed and ma y explain w hy knockdown of YY1 impacts so significantl y on locus folding.

Figure 6 .
Figure 6.Antisense eRNAs encoded by E 3-1 r epr ess YY1 r ecruitment.( A ) Tempor al analysis of E 3-1 tr anscription in PIPER-15 cells.Tr anscription of the E 3-1 enhancer was analysed by RT-qPCR in PIPER-15 cells following induction.Data are normalized to Hprt expression.( B ) Integrator is recruited to both E 3-1 and V 1p in PIPER-15 cells.Integrator binding at the E 3-1 enhancer and V 1 promoter analysed by ChIP-qPCR in PIPER-15 cells following induction.The fold enrichment at E 3-1, V 1p and Intgene III is shown.Binding falls from peaks le v els at 4 hpi but remains above that at 0 hpi; this may be due to Integrator turning over stalled RNAPII ( 86 ), that is transferred to the promoter at later time points.All values are normalized to binding at Intgene III as a negati v e control.( C ) Left: GRO-seq data fr om pr o-B cells was reanalysed using the Galaxy w e b server.Signal peaks of ATAC-seq and ChIP-seq data fr om pr o-B cells map to the central region of the E 3-1 enhancer.Visualization of the mapped reads was performed in IGV.Genomic coordinates of the E 3-1 enhancer are shown.( D ) Temporal analysis of expression of sense (sE 3-1e; upper) and antisense (asE 3-1e; lower) eRNAs by RT-qPCR in PIPER-15 cells following induction.Data are normalized to Hprt expression.( E ) Analysis of the relati v e interaction frequency of Dpn II fragments from the E 3-1 enhancer in PIPER-15 cells expressing scrambled (shSCR), or shRNAs against sense (shsE 3-1e, upper) or antisense (shasE 3-1e, lower) eRNAs.The significance of the difference in interactions is gi v en in Supplementary Figures S6H and S6I. ( F ) YY1 binding to E 3-1 in PIPER-15 cells expressing scrambled (shSCR), sense shsE 3-1e and antisense shasE 3-1e E 3-1 eRNAs.All values are normalized to binding at Intgene III as a negati v e control.( G ) Nati v e agarose gel electrophoresis of sense (sE 3-1e) and antisense (asE 3-1e) eRNAs.H. YY1 binding to sense (sE 3-1e) and antisense (asE 3-1e) eRNAs in PIPER-15 cells expressing scrambled (shSCR) or shRNA against antisense (shasE 3-1e) eRNA.The fold enrichment over input is shown.Error bars show standard error of the mean (SEM) from three biological replicates.